UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Chemometrics and metabolomics of Cannabis sativa L. Mudge, Elizabeth Margaret 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2019_may_mudge_elizabeth.pdf.pdf [ 3.89MB ]
Metadata
JSON: 24-1.0376606.json
JSON-LD: 24-1.0376606-ld.json
RDF/XML (Pretty): 24-1.0376606-rdf.xml
RDF/JSON: 24-1.0376606-rdf.json
Turtle: 24-1.0376606-turtle.txt
N-Triples: 24-1.0376606-rdf-ntriples.txt
Original Record: 24-1.0376606-source.json
Full Text
24-1.0376606-fulltext.txt
Citation
24-1.0376606.ris

Full Text

Chemometrics and Metabolomics of Cannabis sativa L.  by Elizabeth Margaret Mudge  M.Sc., University of Alberta, 2011 B.A.Sc., Dalhousie University, 2008  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE COLLEGE OF GRADUATE STUDIES (Chemistry) THE UNIVERSITY OF BRITISH COLUMBIA (Okanagan) February 2019  © Elizabeth Margaret Mudge, 2019  ii The following individuals certify that they have read, and recommend to the College of Graduate Studies for acceptance, a thesis/dissertation entitled: Chemometrics and Metabolomics of Cannabis sativa L.  submitted by Elizabeth Margaret Mudge   in partial fulfillment of the requirements of   the degree of  Doctor of Philosophy    Dr. Susan Murch, Irving K. Barber School of Arts and Sciences, Department of Chemistry Supervisor Dr. Paula Brown, Irving K. Barber School of Arts and Sciences, Department of Biology Supervisory Committee Member Dr. Wesley Zandberg, Irving K. Barber School of Arts and Sciences, Department of Chemistry Supervisory Committee Member Dr. Philip Ainslie, Faculty of Health and Social Development, School of Health & Exercise Science University Examiner Dr. Cory Harris, Biology, University of Ottawa External Examiner  Additional Committee Members include: Dr. Michael Deyholos, Irving K. Barber School of Arts and Sciences, Department of Biology Supervisory Committee Member    iii Abstract Cannabis is a domesticated crop that has a long history of medical and recreational use. Strains have been selected through informal breeding programs with undisclosed parentage and criteria. The term ‘strain’ refers to minor morphological differences and grower branding rather than distinct stabilized varieties. Anecdotal evidence suggests that there are significant variations in pharmacological effects of different strains, for which the phytochemicals responsible are unknown. The overall objective of this research was to develop analytical methodologies and metabolomic tools to characterize the phytochemical diversity of Cannabis strains within the Canadian marketplace. The expected outcomes were to understand phytochemical relationships caused by domestication and the varying pharmacological effects of the strains. A method was optimized and validated for the quantitation of ten cannabinoids in Cannabis flowers. With nine test materials, the relative standard deviations ranged from 0.78 to 10.08 % with intermediate precision HorRat values of 0.3 to 2.0. In the thirty-three Canadian market samples, the two cannabinoids Δ9-tetrahydrocannabinolic acid and cannabidiolic acid contents ranged from 0.76 to 20.71% and <MDL to 18.11%, respectively. Five clusters of strains were identified based on the ranges of total THC/CBD in the strains. Using targeted-untargeted metabolomics, the relationships between known and unknown cannabinoids were evaluated to identify CBD correlated cannabinoids, and losses in the phytochemical diversity of strains based on the distribution of cannabinoids across strain clusters. Terpenes were determined by headspace profiling in the strains. In total, 67 terpenes were detected and grouped according to their cannabinoid clusters. There were several terpenes identified in unique clusters, which highlights the impacts of aroma and breeding practices on the losses of phytochemical diversity. A data fusion model and post hoc algorithms were used to identify relationships between all cannabinoids and terpenes. There was a strong indication of a unique biosynthetic pathway for the synthesis of terpinolene, nine iv additional monoterpenes and three unidentified cannabinoids. Isolation and characterization of the most prominent cannabinoid was identified as THCA-C4 with a butyl sidechain. Another correlated cannabinoid was identified as THCVA. These two cannabinoids are present in higher quantities in the presence of terpinolene suggesting that breeding for increased cannabinoid production, switches on the biosynthesis of terpinolene in some strains. This research provides novel insight into the impacts of breeding, selection and domestication syndrome in Cannabis. There is a loss of phytochemical diversity in many strains that can potentially impact pharmacological activities.   v Lay Summary There are hundreds of marijuana strains available in the marketplace due to underground breeding that occurred prior to legalization. Breeders selected strains based on the plants’ ability to get them ‘high’ (potency), experience (relaxing, stimulating, etc.), aroma, plant size, and overall yield. Strains available can vary in effectiveness, aroma, and quality for both recreational and medical purposes. This work focused on looking at the in-depth chemical variation of different strains and how these breeding practices have impacted chemical diversity. A collection of thirty-three strains were collected from licensed producers in Canada and evaluated for cannabinoids and terpenes. Using advanced algorithms, it was possible to identify strains which no longer contain certain chemicals and may have lost some unique characteristics. A relationship between different compounds identified a group of strains with a very different chemical composition, which highlighted the impacts of breeding and loss of chemical diversity of marijuana strains. vi Preface A version of Chapter 2 has been published. Mudge EM, Murch SJ, Brown PN. 2017. Leaner and greener analysis of Cannabinoids. Analytical and Bioanalytical Chemistry. 409(12): 3153-3163.  I conducted the study design, optimized and developed the methods, conducted the experiments and analyzed the data. I prepared the graphs, tables and interpreted the results of the data. I wrote the original draft of the paper. A version of Chapter 3 has been published. Mudge EM, Murch SJ, Brown PN. 2018. Chemometric analysis of cannabinoids: Chemotaxonomy and domestication syndrome. Scientific Reports. 8: 13090.  I conducted the majority of the experiments and analyzed all of the data. I reviewed the literature to critically discuss the results. I also summarized and graphed the data and wrote the original draft of the paper. A version of Chapter 4 has been submitted for publication. Mudge EM, Brown PN, Murch SJ. 2019. The terroir of Cannabis: Terpene metabolomics as a tool to understand Cannabis sativa selections. Planta Medica. Submitted. I conducted the majority of the experiments and analyzed all of the data. I reviewed the literature to critically discuss the results. I also summarized and graphed the data and wrote the original draft of the paper. A version of Chapters 5 and 6 has been submitted for publication. vii Mudge EM, Brown PN, Murch SJ. 2019. Use of chemometric models to elucidate relationships between cannabinoids and terpenes in medical Cannabis. Journal of Natural Products. Submitted. I conducted the majority of the experiments and analyzed all of the data. I reviewed the literature to critically discuss the results. I also summarized and graphed the data and wrote the original draft of the paper. Presentations: Data from Chapter 2, 3 and 4 were presented at: Mudge EM, Murch SJ, Brown PN. 2016. Quantitation of cannabinoids in medical marihuana flowers and oils. Halifax, NS. June 6-9, 2016. Canadian Chemistry Society Annual Conference. Mudge EM, Murch SJ, Brown PN. 2016. Evaluating phytochemical variation in medical marihuana. Copenhagen, DK. July 24-28, 2016. Joint Natural Products Conference. Mudge EM, Murch SJ, Brown PN. 2016. Quantitation and variation of cannabinoids in medical Cannabis. Kelowna, BC. September 6, 2016. UBCO Biology Graduate Symposium. Mudge EM, Murch SJ, Brown PN. 2017. Quantitation and variation of cannabinoids in medical Cannabis flowers. Vancouver, BC. May 8-11, 2017. NHPRS Meeting. Mudge EM, Murch SJ, Brown PN. 2017. Chemical approaches to evaluate the diversity of medicinal Cannabis. Toronto, ON. May 29-June 1. Canadian Society of Chemistry Conference. viii Mudge EM, Murch SJ, Brown PN. 2018. Evaluation of terpenes in Cannabis sativa: impacts of human selection on phytochemical diversity. Oral Presentation in Chemistry of Cannabis Symposium at 101st Canadian Chemistry Conference and Exhibition, Edmonton AB.   ix Table of Contents Abstract ..............................................................................................................................iii Lay Summary ...................................................................................................................... v Preface ................................................................................................................................vi Table of Contents ...............................................................................................................ix List of Tables .................................................................................................................... xiii List of Figures ...................................................................................................................xv List of Abbreviations .........................................................................................................xx Acknowledgements ......................................................................................................... xxii Dedication ....................................................................................................................... xxiv Chapter 1: Introduction ...................................................................................................... 1 1.1 Introduction ......................................................................................................... 1 1.2 History of Cannabis utilization ............................................................................. 2 1.2.1 Botany & taxonomy ..................................................................................... 2 1.2.2 Domestication syndrome ............................................................................. 5 1.2.3 Plant development ....................................................................................... 7 1.2.4 Chemistry and regulatory compliance .......................................................... 7 1.3 Phytochemistry ................................................................................................... 8 1.3.1 Cannabinoids .............................................................................................. 9 1.3.2 Terpenes ....................................................................................................11 1.3.3 Flavonoids ..................................................................................................12 1.3.4 Other phytochemicals .................................................................................13 1.4 Analytical assays ...............................................................................................13 1.4.1 Cannabinoids .............................................................................................14 1.4.2 Terpenes ....................................................................................................17 1.4.3 Other phytochemicals .................................................................................18 1.5 Metabolomic evaluations ...................................................................................18 1.5.1 Forensic applications ..................................................................................19 1.5.2 Chemotaxonomy applications .....................................................................20 1.5.3 Strain differentiation ....................................................................................21 1.6 Study rationale and objectives ...........................................................................24 Chapter 2: Single laboratory validation of an HPLC-UV method to quantify cannabinoids in Cannabis ................................................................................................26 2.1 Synopsis ............................................................................................................26 x 2.2 Experimental ......................................................................................................27 2.2.1 Reagents ....................................................................................................27 2.2.2 Calibration standards ..................................................................................27 2.2.3 Test materials .............................................................................................28 2.2.4 HPLC analysis ............................................................................................28 2.2.5 Method optimization ...................................................................................29 2.2.5.1 Fractional factorial ..................................................................................29 2.2.5.2 Analyte stability .......................................................................................29 2.2.6 Preparation of test materials .......................................................................29 2.2.7 Single-laboratory validation parameters ......................................................30 2.2.7.1 Selectivity ...............................................................................................30 2.2.7.2 Linearity ..................................................................................................30 2.2.7.3 Repeatability and intermediate precision .................................................31 2.2.7.4 Recovery ................................................................................................31 2.2.7.5 Limits of detection and quantitation .........................................................32 2.3 Results ..............................................................................................................32 2.3.1 Method optimization ...................................................................................32 2.3.1.1 Extraction of tissues ................................................................................33 2.3.1.2 Chromatographic optimization ................................................................37 2.3.1.3 Analyte stability .......................................................................................39 2.3.2 Method validation .......................................................................................40 2.3.2.1 Selectivity ...............................................................................................40 2.3.2.2 Linearity ..................................................................................................41 2.3.2.3 Repeatability ...........................................................................................42 2.3.2.4 Intermediate precision .............................................................................44 2.3.2.5 Recovery ................................................................................................44 2.3.2.6 Limits of detection and quantitation .........................................................45 2.4 Discussion .........................................................................................................46 Chapter 3: Chemometric Analysis of Cannabinoids: Chemotaxonomy and Domestication Syndrome .................................................................................................48 3.1 Synopsis ............................................................................................................48 3.2 Experimental ......................................................................................................49 3.2.1 Reagents ....................................................................................................49 3.2.2 Test materials .............................................................................................49 xi 3.2.3 Targeted metabolomics of cannabinoids ....................................................50 3.2.4 Untargeted metabolomics ...........................................................................50 3.2.5 Data analysis ..............................................................................................51 3.3 Results ..............................................................................................................51 3.3.1 Targeted metabolomics of cannabinoids ....................................................51 3.3.2 Classification of strains ...............................................................................53 3.3.3 Untargeted metabolomic analysis ...............................................................55 3.3.4 Relationships between known and unknown cannabinoids .........................55 3.3.5 Putative identifications and pathways .........................................................61 3.4 Discussion .........................................................................................................67 Chapter 4: Domestication Syndrome and Metabolomics of Volatile Constituents in Cannabis ............................................................................................................................74 4.1 Synopsis ............................................................................................................74 4.2 Experimental ......................................................................................................75 4.2.1 Test materials .............................................................................................75 4.2.2 Reference standards ..................................................................................75 4.2.3 Evaluation of volatile constituents ...............................................................76 4.2.4 Chemometrics ............................................................................................76 4.2.4.1 Metabolite profiling ..................................................................................76 4.2.4.2 Identification of metabolite relationships ..................................................77 4.2.4.3 Multivariate classification ........................................................................77 4.3 Results ..............................................................................................................77 4.3.1 Terpene profiles..........................................................................................77 4.3.2 Terpene profile by strain .............................................................................80 4.3.3 Aroma characterization ............................................................................. 100 4.3.4 Terpene Metabolomics ............................................................................. 103 4.4 Discussion ....................................................................................................... 106 Chapter 5: Multiblock Data Fusion Model to Identify Relationships Between Cannabinoids and Terpenes .......................................................................................... 117 5.1 Synopsis .......................................................................................................... 117 5.2 Experimental .................................................................................................... 119 5.2.1 Plant materials .......................................................................................... 119 5.2.2 Metabolite profiling ................................................................................... 120 5.2.3 High-level data fusion ............................................................................... 120 xii 5.2.4 Chemometrics .......................................................................................... 120 5.3 Results ............................................................................................................ 121 5.3.1 Interrelationships between terpenes and cannabinoids............................. 121 5.3.2 High-level data fusion of terpenes and cannabinoids ................................ 122 5.3.3 Metabolomic based chemotyping of Cannabis strains .............................. 124 5.3.4 Cannabinoids and terpenes: same strain, different producer comparison . 129 5.3.5 Cannabinoids and terpenes: lot-to-lot comparisons .................................. 133 5.4 Discussion ....................................................................................................... 141 Chapter 6: Untargeted Metabolomics for Discovery of Phytochemical Relationships: Exploration into the Terpinolene Pathway .................................................................... 149 6.1 Synopsis .......................................................................................................... 149 6.2 Experimental .................................................................................................... 150 6.2.1 Test samples ............................................................................................ 150 6.2.2 Reagents and calibration standards.......................................................... 150 6.2.3 Terpinolene quantitation ........................................................................... 150 6.2.3.1 Quantitation validation .......................................................................... 151 6.2.4 CMPD12 isolation ..................................................................................... 152 6.2.5 Structure elucidation ................................................................................. 153 6.2.5.1 NMR spectroscopy................................................................................ 153 6.2.5.2 LC-MS acquisition ................................................................................. 153 6.3 Results ............................................................................................................ 153 6.3.1 Terpene quantitation ................................................................................. 153 6.3.1.1 Terpene quantitation: validation ............................................................ 164 6.3.2 CMPD12 isolation and characterization .................................................... 164 6.3.3 LC-MS characterization of correlated cannabinoid (CMPD7) .................... 170 6.4 Discussion ....................................................................................................... 171 Chapter 7: Conclusion .................................................................................................... 177 References ....................................................................................................................... 183    xiii List of Tables Table 2.1 Concentrations of the cannabinoids used in the mixed calibration standards for the seven-point calibration curves. ............................................................................31 Table 2.2 Runs prepared for the partial factorial design to optimize the sample preparation of cannabinoid extractions. ..................................................................................33 Table 2.3 Cannabinoid stability of mixed standard solution prepared in 80% methanol stored at three different temperatures in the dark. ..........................................................39 Table 2.4 Cannabinoid stability of authentic Cannabis extracts in 80% methanol at room temperature after 24 hours in dark and light conditions.. .....................................40 Table 2.5 Cannabinoid stability in authentic Cannabis extracts in 80% methanol at 4 and 22 °C for 48 hours.. ..................................................................................................40 Table 2.6 Cannabinoid stability in authentic Cannabis extracts in 9:1 v/v methanol:chloroform stored at 4 and 22  °C for 48 hours.....................................40 Table 2.7 Resolution and peak purity for each cannabinoid used to determine method selectivity. ...........................................................................................................41 Table 2.8 Linear regression data for each cannabinoid. ......................................................41 Table 2.9 Repeatability and intermediate precision for cannabinoid quantitation in Cannabis dried flowers. .......................................................................................................42 Table 2.10 Repeatability (% RSD) for the additional cannabinoids detected in Cannabis flowers which were assessed after the original validation was completed. ..........44 Table 2.11 Accuracy for the quantitation of major cannabinoids using spike recovery and stinging nettle as the matrix blank. ......................................................................45 Table 2.12 Method detection limit and limit of quantitation of cannabinoids in solution and their respective concentrations in dried flowers using the EPA MDL procedures. 45 Table 3.1 Strains of Cannabis were clustered into 5 distinct groups that could be separated by the flow of metabolites through the CBD and THC pathways. .........................53 Table 3.2 Pearson correlation coefficients of all cannabinoids relative to the four major cannabinoids (THCA, CBDA, THC and CBD). .....................................................60 xiv Table 3.3 UV spectrum of known and unknown cannabinoids along with their putative identification based on UV spectrum, elution order and correlations. ...................71 Table 4.1 Identification of terpenes in Cannabis by reference standard mix (std. mix), mass spectral and retention index comparisons............................................................78 Table 4.2 Aroma descriptors for each of the terpenes identified within the Cannabis strains, grouped based on their presence within the cannabinoid classes. .................... 102 Table 4.3 Pharmacological activities of monoterpenes and sesquiterpenes identified in Cannabis. .......................................................................................................... 110 Table 5.1 Results of the significance analysis of microarray to identify the cannabinoids and terpenes significant in the chemotaxonomic classification of Cannabis strains. . 126 Table 5.2 Results of the 1D-self organizing map (1DSOM) for identifying the metabolite clusters responsible for the chemotaxonomic classification of Cannabis. .......... 128 Table 5.3 Summary of the between strain variance of two strains purchased from different producers. ......................................................................................................... 130 Table 5.4 Variance of cannabinoids observed between two lots of the same strain from the same producer. ................................................................................................. 133 Table 5.5 Variance of terpene profiles between two lots of the same strain from the same producer. ........................................................................................................... 137 Table 6.1 Total monoterpene and sesquiterpene concentrations measured in all thirty-three Cannabis strains................................................................................................ 155 Table 6.2 Carbon (δC) and proton (δH) NMR spectroscopic data (400 MHz, CDCl3) for CMPD12, identified as THCA-C4. ..................................................................... 168   xv List of Figures Figure 1.1 Total number of publications related to Cannabis research recorded in Web of ScienceTM as of September 20, 2018 ................................................................... 1 Figure 1.2 Trimmed, dried female Cannabis sativa inflorescences. ...................................... 3 Figure 1.3 Structures of the base structure of the nine cannabinoids classes found in Cannabis sativa L. ................................................................................................ 9 Figure 1.4 Structures of the major monoterpenes and sesquiterpenes commonly detected in Cannabis flowers. ................................................................................................12 Figure 2.1 Variation of the total cannabinoid content determined using the partial factorial design to optimize the sample preparation. .........................................................33 Figure 2.2 Cannabinoid content of CBDA, CBD, THCA and THC evaluated using two different extraction solvents, 9:1 v/v methanol:chloroform and 80% methanol, and two different sample masses, 100 and 200 mg.. ..................................................34 Figure 2.3 Overlay of the chromatographic separation of cannabinoids extracted with 9:1 methanol:chloroform (black) and 80% methanol (red). ........................................35 Figure 2.4 Comparison of a single extraction versus a 2 times extraction to confirm extraction effectiveness.. .....................................................................................35 Figure 2.5 Cannabinoid content of CBDA, CBD, THCA and THC after sonication for 15, 30 and 60 minutes. ...................................................................................................36 Figure 2.6 Cannabinoid content of CBDA, CBD, THCA and THC after sonication for 5, 10, 15 minutes and 15 minutes with vortexing every 5 minutes.. ...............................37 Figure 2.7 Chromatographic separation of cannabinoids using a Kinetex C18 column (100 x 2.1 mm, 1.8 µm) with gradient elution at 220 nm. (A) mixture of cannabinoids standards (B) authentic Cannabis extract. ...........................................................38 Figure 3.1 Biosynthetic pathway of cannabinoids originating from olivetolic acid and geranyl pyrophosphate. ...................................................................................................52 Figure 3.2 Biosynthetic pathway of cannabinoids originating from divarinolic acid and geranyl pyrophosphate.. ......................................................................................54 xvi Figure 3.3 PCA scores plot of cannabinoid profiles classified according to THC/CBD contents. .............................................................................................................56 Figure 3.4 PCA loadings plot of cannabinoid profiles classified according to THC/CBD contents. .............................................................................................................56 Figure 3.5 Multiple linear regression model to estimate the THCA content from the entire cannabinoid dataset. ...........................................................................................57 Figure 3.6 Improved MLR model to estimate the THCA content from a reduced data set of 14 cannabinoids. .................................................................................................58 Figure 3.7 MLR model to estimate CBDA content from the entire cannabinoid dataset. ......58 Figure 3.8 Improved MLR model to estimate the CBDA content from a reduced dataset of 14 cannabinoids. ......................................................................................................59 Figure 3.9 Heatmap illustrating the Pearson correlation coefficients for all of the cannabinoids detected using HPLC-UV separation. ............................................61 Figure 3.10 Cannabinoid content of the unidentified cannabinoids found throughout the sample set. (a) CMPD4, (b) CMPD7, (c) CMPD8, (d) CMPD9, (e) CMPD10, (f) CMPD11, (g) CMPD14, (h) CMPD16, (i) CMPD19, (j) CMPD21. .........................64 Figure 3.11 Cannabinoid content of the unidentified cannabinoids found in CBD-rich strains. (a) CMPD1, (b) CMPD3, (c) CMPD5, (d) CMPD6, (e) CMPD15, (f) CMPD18. ....66 Figure 3.12 Unidentified cannabinoids found only in THC-dominant strains. (a) CMPD2, (b) CMPD12, (c) CMPD20. .......................................................................................67 Figure 3.13 UV spectra of acidic and neutral unidentified cannabinoids in comparison to known cannabinoids. (a) CMDP1, (b) CBDA, (c) CMPD18, (d) CBD. ..................70 Figure 4.1 Terpene profiles identified as those present across the entire dataset. (a) camphene (b) β-caryophyllene (c) guaia-3,9-diene (d) α-guaiene (e) humulene (f) D-limonene (g) β-maaliene (h) β-myrcene (i) α-pinene (j) β-pinene (k) selina-3,7(11)-diene and (l) valencene. ..........................................................................83 Figure 4.2 Terpene profiles identified as those present across the different cannabinoids classes, but not present in all strains. (a) endo-borneol (b) camphene hydrate (c) copaene (d) β-cubebene (e) exo-fenchol (f) fenchone (g) Germacrene A (h) α-xvii gurjunene derivative (i) γ-gurjunene (j) γ-muurolene (k) trans-2-pinanol (l) z-sabinine hydrate (m) 4,11-selinadiene (n) α-selinene (o) β-selinene (p) α-terpineol (q) Ylangene. .......................................................................................................88 Figure 4.3 Pearson correlations between monoterpenes and sesquiterpenes within the Cannabis dataset. ...............................................................................................89 Figure 4.4 Terpene profiles present primarily in THC-dominant strains. (a) α-amorphene (b) 2-carene (c) caryophyllene oxide (d) α-cubenene (e) β-elemene (f) γ-elemene (g) (Z,Z)-α-farnesene (h) germacrene B (i) β-sesquiphellandrene. ............................92 Figure 4.5 Terpene profiles for terpenes found predominantly in mid-level THC-dominant strains. (a) δ-cadiene (b) α-gurjunene (c) santolina triene (d) sesquiterp-1 (unidentified). ......................................................................................................93 Figure 4.6 Terpene profiles representing a unique group of terpenes that dominate both THC-dominant and CBD-THC hybrid strains. (a) α-bulnesene (b) bulnesol (c) 3-carene (d) p-cymene (e) p-cymenene (f) α-eudesmol (g) cis-β-farnesene (h) -fenchene (i) linalool (j) α-phellandrene (k) β-phellandrene (l) α-santolene (m) δ-selinene (n) α-terpinene (o) γ-terpinene (p) terpinen-4-ol (q) terpinolene (r) α-thujene. ...............................................................................................................98 Figure 4.7 Terpenes predominantly found in higher CBD strains. (a) alloaromadendrene (b) cis-α-bisabolene (c) 10-epi-γ-eudesmol (d) guaiol (e) cis-β-ocimene (f) trans-β-ocimene (g) sabinene. ....................................................................................... 100 Figure 4.8 Principal component analysis (PCA) of the monoterpene and sesquiterpene profiles for the Cannabis dataset ....................................................................... 104 Figure 4.9 PCA loadings plot of PC 1 and PC2 describing the influence of terpenes on the variation of Cannabis strains in Figure 4.8. ........................................................ 104 Figure 4.10 PCA of the monoterpene and sesquiterpene profiles within the Cannabis dataset after implementing a data reduction strategy. ....................................... 105 Figure 4.11 PCA loadings plot for monoterpenes and sesquiterpenes after implementation of the data reduction strategy to identify the metabolites influencing the strain clustering of the PCA scores plot in Figure 4.10. ............................................... 106 xviii Figure 5.1 Biosynthetic pathway of geranyl pyrophospate via the MEP pathway and olivetolic acid via the polyketide synthesis pathway to produce the precursors for monoterpene and cannabinoid biosynthesis. ..................................................... 118 Figure 5.2 Pearson correlation heatmap describing correlations between cannabinoids and terpenes. ........................................................................................................... 122 Figure 5.3 Unsupervised data fusion clustering with a multiple factor analysis (MFA) factor plot describing dimensions 1 and 2 of the cannabinoid, monoterpene and sesquiterpene profiles within the Cannabis dataset. .......................................... 123 Figure 5.4 Multiple factor analysis correlation plot of the cannabinoids, monoterpenes and sesquiterpenes within the Cannabis dataset. ..................................................... 123 Figure 5.5 Hierarchical cluster analysis of the strains based on their cannabinoid and terpene compositions. ....................................................................................... 125 Figure 5.6 Decision tree built from the HCA clusters using different cannabinoids and terpenes to identify the appropriate cluster. ....................................................... 129 Figure 5.7 Bar graph comparing the content of each cannabinoid detected the nordle samples provided by two different producers. .................................................... 131 Figure 5.8 Terpene headspace profiles comparing the nordle samples from two different producers. ......................................................................................................... 132 Figure 5.9 Cannabinoid quantitative profiles comparing the lot-to-lot variance of nine Cannabis strains obtained two years apart. (a) SensiStar, (b) Ice cream, (c) Wappa, (d) Nebula CBD, (e) Spoetnik, (f) Chronic Thunder, (g) Acapulco Gold, (h) Blueberry Lambsbread, (i) Bubba Kush. ............................................................ 136 Figure 5.10 GC headspace data of the monoterpene and sesquiterpene profiles of two lots of wappa which had the lowest variance of the nine strains evaluated. ............. 139 Figure 5.11 GC headspace data of the monoterpene and sesquiterpene profiles of two lots of Acapulco gold which had the highest variance of the nine strains evaluated. 140 Figure 5.12 Proposed pathway based on linked cannabinoid and terpenes for the biosynthesis of terpinolene and other correlated terpenes. ................................ 146 xix Figure 6.1 Correlation plots evaluating the relationship between total cannabinoids content and terpenes. (a) Correlation plot of total cannabinoid content versus total monoterpene content and (b) correlation plot of total cannabinoid content versus total sesquiterpene content. .............................................................................. 156 Figure 6.2 Monoterpene contents of (a) α-pinene, (b) β-myrcene, (c) limonene and (d) terpinolene across the Cannabis dataset. .......................................................... 158 Figure 6.3 Monoterpene profiles of the six terpinolene-dominant strains (a) can16, (b) can17, (c) can19, (d) can21 (e) can32, (f) can 33 in comparison to three representative non-terpinolene-dominant strains (g) can39, (h) can22, (i) can18. .......................................................................................................................... 160 Figure 6.4 Sesquiterpene profiles of the six terpinolene-dominant strains (a) can16, (b) can17, (c) can19, (d) can21 (e) can32, (f) can 33 in comparison to three representative non-terpinolene-dominant strains (g) can39, (h) can22, (i) can18. .......................................................................................................................... 163 Figure 6.5 Semi-preparative HPLC separation of the cannabinoid fraction obtained after flash chromatography to isolate CMPD12 at 220 nm. ........................................ 165 Figure 6.6 Final semi-preparative HPLC separation of CMPD12 from other minor contaminants at 220nm to isolate CMPD12. ...................................................... 166 Figure 6.7 Product ion spectra from (a) THCA with a molecular ion of m/z 359 and (b) THCA-C4 (CMPD12) with a molecular ion of m/z 345. ...................................... 169 Figure 6.8 Chromatographic separation of CMPD12 (THCA-C4; blue) and THCA (red) with detection at 220 nm. .......................................................................................... 170 Figure 6.9 Structure of CMPD12, identified as THCA-C4 with a butyl sidechain on the polyketide moiety............................................................................................... 170 Figure 6.10 Product ion scan of THCVA with a molecular ion at m/z 331 used to confirm the identity of CMPD7. ............................................................................................ 171    xx List of Abbreviations ACS  American Chemical Society AOAC  AOAC International C18  Octadecyl Carbon Chain  C21  Henicosyl Carbon Chain CBC  Cannabichromene CBD  Cannabidiol  CBDA  Cannabidiolic Acid CBDM  Cannabidiol Monomethyl Ether CBDMA Cannabidiol Monomethyl Ether Acid CBDV   Cannabidivarin  CBDVA Cannabindivarinic Acid  CBG  Cannabigerol  CBGA  Cannabigerolic Acid CBGVA Cannabigerovarin Acid  CBL  Cannabicyclol  CBN  Cannabinol CDC  Center for Disease Control CDCl3  Deuterated chloroform CHCl3   Chloroform CMPD  Compound (Unidentified Cannabinoid Label) COSY  Correlation Spectroscopy  CSTPS6TN Gene Encoding Terpene Synthase for cis-β-ocimene DAD  Diode Array Detector DB-5  5% Phenyl 95% Dimethylarylene Siloxane DL  Detection Limit DMAPP Dimethylallyl Pyrophosphate DNA  Deoxyribonucleic Acid  1DSOM One Dimensional Self-Organizing Map EPA  Environmental Protection Agency  EU  European Union FDR  False Discovery Rate FID  Flame Ionization Detection FPP  Farnesyl Pyrophosphate  g  Relative Centrifugal Force (G-force) GC  Gas Chromatography GOT  Geranyl Diphosphate: Olivetolate Geranyltransferase GPP  Geranyl Pyrophosphate HCA  Hierarchical Cluster Analysis HMBC  Heteronuclear Multiple-Bond Correlation Spectroscopy  HPLC  High Performance Liquid Chromatography HSQC  Heteronuclear Single Quantum Coherence Spectroscopy ID  Identification IPP  Isopentenyl Pyrophosphate LC  Liquid Chromatography LOQ  Limit of Quantitation MΩ  Megaohm MDL  Method Detection Limit MCR  Multiple Curve Resolution xxi MeOH  Methanol  MEP  Methylerythritol 4-Phosphate Pathway MFA  Multiple Factor Analysis MLR  Multiple Linear Regression  MS  Mass Spectrometry MVA  Mevalonic Acid Pathway  NIST  National Institute of Standards and Technology NMR  Nuclear Magnetic Resonance Spectroscopy NPP  Neryl Pyrophosphate PC  Principal Component  PCA  Principal Component Analysis PLS-DA Partial Least Squares – Discriminant Analysis psi  Pounds per Square Inch PTFE  Polytetrafluoroethylene (Teflon) QCRIT  Q-critical QEXP  Q-experimental r2   Coefficient of Variation RRLC  Rapid Resolution Liquid Chromatography  RSD  Relative Standard Deviation  SAM  Significance Analysis of Microarrays SEM  Standard Error of the Mean SFC  Supercritical Fluid Chromatography std  Standard Δ8-THC delta-8 Tetrahydrocannabinol  Δ9-THC delta-9 Tetrahydrocannabinol  THC  Tetrahydrocannabinol THCA  Tetrahydrocannabinolic Acid THCA-C4  Tetrahydrocannibinolic Acid – Butyl  THCV  Tetrahydrocannabivarin  THCVA Tetrahydrocannabivarinic Acid TLC  Thin Layer Chromatography TMS  Trimethylsilane TPS  Terpene Synthase TSP  Trimethylsilylpropanoic acid UNODC United Nations Office on Drugs and Crime UV  Ultraviolet v/v  Volume by Volume w/w  Weight by Weight  xxii Acknowledgements I am forever grateful for the opportunities that have been provided to me through my PhD studies. I would like to thank my advisors, Dr. Paula Brown and Dr. Susan Murch for their support, direction and dedication to this research. Their feedback and encouragement allowed me to become a stronger and more confident researcher. Thank you for expanding my knowledge, understanding and experience in plant metabolomics and for providing much needed guidance throughout my studies. I am grateful for your unwavering support throughout this journey. I would also like to thank my supervisory committee members, Dr. Michael Deyholos and Dr. Wesley Zandberg for your sound advice and encouragement throughout my graduate studies.  I would also like to acknowledge the support of my lab mates at BCIT. Hong Sy, Hazrah Moothoo, Dr. Michael Chan, Ronan Yu, Jamie Finley, Dr. Ying Liu, Dr. Xiaohui Zhang and Rebecca Robertson. In particular, I would like to thank Hazrah for your assistance in the lab, Michael for your constant advice and NMR support, Jamie for being open to discuss chemometrics and Ying and Hong for always being supportive. Thank you all for supporting this journey, providing advice and being there when I needed it most. Thank you to my fellow Murch labmates; Stephanie Bishop, Jensen Lund and Fiona Tymm for your great conversations, support and advice. I wish you all a prosperous future. I would also like to acknowledge the support of those who encouraged me to pursue my PhD throughout my career; Dr. Andreas Schieber, Dr. Gianfranco Mazzanti, Dr. Joseph Betz, Dr. Catherine Rimmer, Dr. Melissa Phillips and Dr. Jimmy Yuk. I would also like to thank the journal editors, peer reviewers, conference and session chairs for helping disseminate this research. And lastly, but most importantly, I want to thank my friends and family for everything you have done the last three years. Wade, your love, support and encouragement has been xxiii unwavering throughout the last few years and I wouldn’t have even started this if it weren’t for you. Dr. Lauren Borja and Melissa Woodward, thank you so much for your encouragement and support when I needed it most. Rebecca, Sarah, Nathan and Sylvia, thank you for everything. You are the best family I could ask for.   xxiv Dedication I dedicate this thesis to my family. Thank you for your unwavering support, love and encouragement.   1 Chapter 1: Introduction 1.1 Introduction The medicinal use of Cannabis is currently one of the most popular and controversial topics in the scientific literature and public discussion. A simple Google Scholar search of the term “Cannabis” returns more than 525,000 results (September 17, 2018) indicating the scope, popularity and diversity of the available information.  A Web of Science™ search of the scientific literature reports more than 13, 583 studies including 1297 reviews in the last decade with a steadily increasing rate of publication (Figure 1.1). The scientific study of Cannabis spans many disciplines including: psychiatry (32.6 %), substance abuse (23.7 %), neuroscience (11.6 %), psychology (13.8 %) pharmacology & pharmacy (12.4 %), public, environmental and occupational health (5.6 %), general internal medicine (4.6 %), chemistry (3.2 %), toxicology (2.1%), pediatrics (1.9 %), plant sciences (1.9 %) and legal medicine (1.8 %). Interestingly, the plant chemistry and plant physiology are less well studied than the human responses to the phytochemistry.  Figure 1.1 Total number of publications related to Cannabis research recorded in Web of ScienceTM as of September 20, 2018 02004006008001000120014001600180020002008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018Number of PublicationsYear2 1.2 History of Cannabis utilization 1.2.1 Botany & taxonomy The family Cannabaceae consists of approximately 10 genera and 170 species (McPartland, 2018). The most commonly known genera are Humulus (hops), Celtis (hackberry/nettle) and Cannabis (marijuana/hemp). The genus Cannabis has 3 species Cannabis sativa L., Cannabis indica Lam. and the ancient wild species Cannabis ruderalis Jan. The taxonomic classification of these genera continues to be debated since cross breeding has blurred the lines between species, ecotypes and varieties (Small, 2015; McPartland, 2018).  Cannabis is a tall, dioecious annual herb with a distinct leaf structure. The leaves are uneven in size and cluster with 3 to 11 leaflets on the same stem. The leaflet in the middle is typically the longest with the length decreasing as it gets closer to the stem. The leaves are usually lanceolate with margin serrulation, strigose and contain glandular trichomes on the surface (Upton et al., 2014). The height and leaf size vary significantly across the genus. Cannabis sativa plants are typically taller, upwards of 6 meters, while Cannabis indica are around 1-3 meters in height (Clarke, 1981; Upton et al., 2014). The plants can grow to maturation in 4-6 months in outdoor cultivation, but this can be reduced during indoor cultivation where up to 6 crops can be harvested per year (McLaren et al., 2008). Female (pistillate) inflorescences can be varying shades of green and occasionally red and purple, which can be streaked or saturated (Upton et al., 2014). When dried the colors can also be tan to brownish. The flowers are present in dense clusters containing trichome covered bracts and bracteoles (Upton et al, 2014; Clark, 1981). The bracteole (calyx) is composed of two sepals producing a conically shaped sheath and, if pollinated, the seeds (Upton et al., 2014). In most cases, Cannabis used for medical and illicit markets are 3 sensimilla, which are seedless flowers (Small, 2015). Protruding from the calyx are the styles and stigmas (pistils), which can range in color from pale yellow to dark brown (Upton et al., 2014). A figure of a Cannabis flower is shown in Figure 1.2.  Figure 1.2 Trimmed, dried female Cannabis sativa inflorescences. Several features are noted (1) trimmed bracts (2) petiole (3) bracteole (calyx) (4) pistils (5) glandular trichomes.  Several plants have similar morphological characteristics to Cannabis including Hibiscus cannabinus, Acer palmatum and Urtica cannabina, but are easily distinguished based on macroscopic and microscopic characteristics (UNODC, 2009). In the case of medicinal Cannabis, the plants are typically grown in controlled environments, for which identification can easily be confirmed. For seized illicit plants positive identification must be established for legal implications (UNODC, 2009). The flower shapes and the presence of trichomes are distinguishing features. Cystolithic trichomes on the upper surfaces of leaves and glandular trichomes and sessile glands on the lower surfaces are unique to Cannabis 1 3 2 4 5 4 (UNODC, 2009). Organoleptic properties of Cannabis are quite characteristic and due to the wide variety of Cannabis strains available in today’s market, the aromatic descriptions can be highly variable. Some of the descriptors used include: fruity, aromatic, euphoric, spicy, citrus, musty, skunky, floral, sour, diesel, vanilla, blueberry, pineapple, lemon, piney, etc. (Upton et al., 2014). The taste is described as “bitter, resinous, sticky and pungent” (Upton et al., 2014).  The long relationship with humans has hindered agreement on the taxonomic classification of Cannabis species. Cannabis sativa L. was originally classified by Linnaeus in 1753 (Small 2015; Shultes et al., 1975). This species was identified from the European cultivated hemp, used primarily for its fibers. In 1783, Lamarck identified a second species C. indica Lam., which originated in India and was known to be smaller, more highly branched and was used for its resin, strong odor and narcotic properties (Schultes, 1975). In 1924, Russian scientists identified a third species as C. ruderalis Jan. representing wild, undomesticated Cannabis (Janischevsky, 1924). Although somewhat accepted by modern taxonomists, there is speculation that C. ruderalis no longer exists because of human influences and cross breeding (Clarke & Merlin, 2016a).   The major debate regarding Cannabis classification is whether this is a monotypic or polytypic genus. Small et al. (1976) scored 232 diverse populations of Cannabis grown under similar conditions for 47 different attributes and were unable to classify them into two distinct groups (C. sativa and C. indica) and postulated that this is a monotypic genus composed of only C. sativa, while the species itself is quite diverse. More recent investigations using DNA accessions identified three distinct classifications (1) ‘sativa’ related to hemp fiber and seed landraces and European ruderal species (2) ‘indica’ related to far Eastern fiber and seed varieties, narrow-leaf drug strains from Asia, Africa and South America and wide-leaf drug strains from Afghanistan, Pakistan and feral populations from Nepal and India, and (3) ‘ruderal’ accessions from central Asia (Hillig, 2005). These were grouped into seven putative 5 taxa. They concluded that there is an indication of an ancient split between sativa and indica that “may pre-date human intervention” (Hillig, 2005). To further complicate these matters, Cannabis marketing generally describes ‘sativa’ strains as stimulating or uplifting, while ‘indica’ strains are relaxing and sedative. There is limited evidence that these descriptors are related to species or lineage (Sawler et al., 2015). The term strain is used in throughout the Cannabis industry and literature considered to be synonymous to cultivar (Small, 2015; McPartland & Guy, 2017). Strain names are not recognized by botanical nomenclature and many strains do not meet registered cultivar standards, therefore the term describes unregistered cultivars to brand the products (Pollio, 2016; McPartland & Guy, 2017). While hemp, consistently noted as C. sativa ancestry, and marijuana strains, which are commonly noted as C. sativa, C. indica or mixed landraces, have considerable genetic variation, where genotyping was only in partial agreement with ancestral designations (van Bakel et al., 2011; Sawler et al., 2015). With moderate separation observed in the principal component analysis (PCA) plot of the genetic data between marijuana C. sativa and C. indica strains, there is a strong possibility of separate pools of genetic diversity (Sawler et al., 2015). There were many identically named strains that had high genetic variability and hemp cultivars were more strongly correlated with the genetic structure of C. indica (Piluzza et al., 2013; Sawler et al., 2015). 1.2.2 Domestication syndrome Domestication syndrome is described by the variation in the traits of a plant species relative to its wild ancestors prior to human intervention (McKey et al., 2010; Meyer et al, 2012). The traits may include increased fruit or seed size, changes in organoleptic properties (sweetness, flavor, aroma, etc.), changes in phenotypic expression such as size, degree of branching and seed retention. Other traits include changes in reproduction, secondary metabolite changes for which all of these represent a loss of genetic diversity. There are many 6 benefits to domestication including increased crop yield, palatability and reduced labor but undesirable effects include decreased pest resistance or the inability to survive outside cultivation.  The domestication of Cannabis has included human selection, inbreeding and cross breeding as well as natural outcrossing and genome mixing (Clarke & Merlin, 2016b). Strains are not easily delineated by genotype and only moderate correlations have been observed between C. indica and C. sativa ancestry. In addition, large genetic variance has been observed within identically named strains (Sawler et al., 2015; Soler et al., 2017). Standardized, highly controlled programs to breed elite varieties or cultivars by selection of phytochemical profile have been limited (Small, 2015; de Meijer, 2014). It is estimated that there are several hundred or perhaps thousands of strains of Cannabis currently being cultivated in legal and illegal markets worldwide (Small, 2015).  It is possible that chemically identical or very closely related plant material is being sold under several different names by different producers and there is no clear definition of the concept of a “strain”.    Further human interactions have been implemented in Cannabis production. Cannabis producing facilities do not grow male plants for several reasons: elimination of pollination which would cause seed production, it is a more efficient use of space and propagation can be performed with mother plants therefore removing the need to propagate from seeds (Small, 2015). Mother plants are used to clone and propagate a specific strain to ensure clones are genetically identical to one another and all female (Clarke & Merlin, 2016b). The strains will not be susceptible to hybridization as there are no male plants for pollination. Plants can also be propagated through tissue culture in large production operations to maintain genetics (Wang et al., 2009)  7 1.2.3 Plant development Biosynthesis impacts the phytochemical composition of the flowers at harvest. Understanding when flower maturation has completed should result in consistent phytochemical content between batches. Cannabinoids and terpenes are synthesized and stored in the glandular trichomes on Cannabis inflorescences, while minor levels of cannabinoids have been found in leaf trichomes (Mahlberg & Kim, 2004; Turner et al., 1978). Cannabinoids are present in low levels in developing plants, which increase during bract formation; the later stages of plant development (Turner et al., 1981a; Turner et al., 1981b; Aizpurua-Olaizola et al., 2016). 1.2.4 Chemistry and regulatory compliance  Cannabis is cultivated in many countries worldwide for fiber and food purposes from hemp plants. Cultivation is typically controlled by government issued licenses. Canada and several US states have implemented hemp regulations where the content of Δ9-tetrahydrocannabinol (THC) must be less than 0.3 % w/w in any part of the plant (Colorado 2013; Health Canada, 1998; California Senate Bill 566). Regulations in the EU are tighter at 0.2 % w/w THC (EU Commission 2000). Additionally, in Canada any hemp products such as flour or oil must contain less than 10 µg/g THC in finished products. These regulations are to restrict diversion into the illicit market.   Regulations of the sale and use of medical and recreational Cannabis have changed drastically in North America in the past last 10 years. As of September 2018, 31 states in the United States and the District of Columbia (DC) have legalized medical Cannabis, while it is legal for recreational use in nine states and DC. Canada has produced medical Cannabis legally since 2001 which was originally supplied by a single manufacturer, where production was opened to licensed producers in 2013 and has grown to 116 approved producers as of 8 September 2018. On October 17, 2018 Cannabis was legalized for recreational purposes across the country and is regulated similar to alcohol by most provinces (Health Canada, 2018). Quality control measures for both medical and recreational products focus on the major cannabinoids and contaminants such as pesticides, microbial load and solvent resides in extracted products. In Canada, the content of Δ9-tetrahydrocannabinolic acid (THCA), Δ9-tetrahydrocannabinol (THC), cannabidiolic acid (CBDA) and cannabidiol (CBD) must be quantified for each batch and specified on the product labels in both the outgoing medical regulations, Access to Cannabis for Medical Purposes Regulations, and the newly implemented Cannabis Regulations that came into effect October 17, 2018 (Health Canada, 2016; Health Canada, 2018). The supplier for medicinal product in the Netherlands (Bedrocan) quantifies THC, CBD and CBN (Cannabis Bureau, 2004). Washington State requires quantitation of THCA, THC and CBD, while New York State requires defining total THC and total CBD, in addition to reporting nine specific cannabinoids and any others above 0.1 % (Washington State, 2014; New York State, 2015).  1.3 Phytochemistry Cannabinoids are unique phytochemicals found only in Cannabis spp. (Figure 1.3). Minor amounts of cannabinoid-like compounds have been found in only two other plant species, Radula variabilis and Helichrysum sp. (Toyota et al, 2002; Bohlmann & Hoffman, 1979). Over 100 cannabinoids have been discovered in Cannabis, many with biological activities including psychoactivity, seizure-reduction, sedation, nausea reduction, pain reduction and appetite stimulation (Porter et al., 2013; Robson, 2001; Tramer, 2001; Berry, 2002; Russo et al., 2007). Many of the studies on cannabinoid activity are not published in scientific literature, are anecdotal or lack sufficient replication limiting Cannabis’ acceptance as a medicinal product (Lamarine, 2012). Other secondary metabolites in Cannabis include 9 terpenes, flavonoids, phenolics, sterols and carotenoids. In total, more than 500 compounds have been isolated and elucidated from Cannabis spp. (ElSholy & Gul, 2014).   Figure 1.3 Structures of the base structure of the nine cannabinoids classes found in Cannabis sativa L.  1.3.1 Cannabinoids  There are 11 classes of cannabinoids: Δ9-tetrahydrocannabinol type, cannabidiol type, cannabigerol type, cannabichromene type, Δ8-tetrahydrocannabinol type, cannabicyclol type, cannabielsoin type, cannabinol type, cannabinodiol type, cannabitriol type and miscellaneous type (Turner et al., 1980; ElSohly & Slade, 2005; ElSohly & Gul, 2014).  The quantities of major cannabinoids can represent upwards of 10-20% of the dried plant material by weight, while minor cannabinoids are typically less than 1% (Mehmedic et al., 2010). Cannabinoids are prenylated polyketides derived from olivetolic acid (polyketide) and geranyl pyrophosphate (GPP) containing C21 backbones. Cannabinoids naturally occur predominantly in their acid forms in the plants, which decarboxylate to neutral forms during harvest, drying and storage (De Backer, 2012). Decarboxylation of cannabinoids upon heating OHHOH OHHOHHOOHOHOOHHOOHOOHHOOHOHHHOOHOHOOH 10 or smoking allows for direct inhalation and bioavailability of the cannabinoids, while ingestion of THC produces several metabolites including 11-hydroxy-THC, which take longer to be metabolized and secreted from the body (Huestis & Smith, 2007).   Δ9-tetrahydrocannabinol (THC) type cannabinoids are the major cannabinoids present in most drug-type Cannabis strains, which was first isolated by Gaoni and Mechoulam (1964). THC is known primarily for its psychoactivity, but pharmaceutical preparations such as Dronabinol have been used for treatment of nausea and vomiting for chemotherapy patients and for appetite stimulation for patients with AIDS (Plasse, 1991). There are a total of 21 known THC-type cannabinoids. Total THC can be over 20% by weight in dried flowers (Mehmedic et al., 2010).  Cannabidiol (CBD) type cannabinoids are the second most studied cannabinoids in Cannabis due to their interaction with the endocannabinoid system and high abundance in certain strains. Cannabidiol has been shown to reduce the negative side-effects from THC such as irritability and anxiety, while potentiating the benefits of THC such as pain reduction (Russo & Guy, 2006). CBD is also known for its anxiety reduction and anti-inflammatory activity (Mechoulam et al., 2007). There are eight known CBD type cannabinoids which can be present in Cannabis at levels up to 10% (ElSohly & Gul, 2014; Mehmedic et al., 2010). Evidence has shown that high-CBD strains of Cannabis can reduce seizures (Porter & Jacobson, 2013; Maa & Figi, 2014; Devinsky et al., 2014).  Cannabinol (CBN) type cannabinoids are artifacts in Cannabis flowers (Turner et al., 1980). They are degradation products of THC and accumulate in materials after prolonged storage. The cannabinoid class containing cannabigerol acid (CBGA) are the precursors to THCA and cannabidiol acid (CBDA) in cannabinoid biosynthesis (Flores-Sanchez & Verpoorte, 2008). This class is not present in high concentrations in drug-type Cannabis, but can be a major cannabinoid in fiber type cannabinoids that are lacking the THCA or CBDA 11 synthase (Fournier et al., 1987; Aizpurua-Olaizola et al., 2016). Cannabichromene (CBC) type cannabinoids are typically in low abundances of 3 to 8% of the total cannabinoid content, with decreasing content as the plants mature (de Meijer et al., 2009). 1.3.2 Terpenes Terpenes are the volatile constituents that contribute to the distinct odor of Cannabis. Monoterpenes and sesquiterpenes make up the majority of the identified terpenes, while a few di- and tri-terpenes have also been identified. Over 140 terpenes have been identified in different strains of Cannabis, but the composition of different species and strains varies considerably (Giese et al., 2015). Terpene content can range from 0.05 to 1.5% of the dried plant material by weight (Giese et al., 2015).  Although the content of terpenes varies considerably between strains, there are several which are commonly observed. Monoterpenes have been shown to represent 48 to 92% of the terpene content, while the sesquiterpene content ranges from 5 to 49% (Mediavilla & Steinemann, 1997). Monoterpenes commonly found in Cannabis include: β-myrcene, limonene, linalool, α-pinene, β-pinene, α-terpinolene, trans-β-ocimene, camphene, fenchol, and α-terpineol (Mediavilla & Steinemann, 1997; Casano et al., 2011; Hillig, 2004; Ross & ElSohly, 1996). Sesquiterpenes commonly found in Cannabis include: β-caryophyllene, α-humulene, guaiol, trans-β-farnesene and β-eudesmol (Mediavilla & Steinemann, 1997; Casano et al., 2011; Hillig, 2004; Ross & ElSohly, 1996). The structures of many of the major monoterpenes and sesquiterpenes identified in Cannabis are shown in Figure 1.4. 12  Figure 1.4 Structures of the major monoterpenes and sesquiterpenes commonly detected in Cannabis flowers.   Interest in terpenes is growing due to their potential synergistic effects with cannabinoids and as chemotaxonomic markers to classify Cannabis strains (McPartland & Russo, 2001; Russo, 2011; Hillig, 2004; Elzinga et al., 2015). Terpenes are impacted by drying conditions, where a significant loss of monoterpenes occurs during drying and storage, while loss of sesquiterpenes is less affected (Ross & ElSohly, 1996). Sensory evaluation of Cannabis essential oil aromas varied considerably where oils high in sesquiterpenes were considered bad smelling, while those high in monoterpenes received a good rating (Mediavilla & Steinemann, 1997). Odor descriptors were not provided in this study, but this preliminary data suggests that long term storage may result in losses of monoterpenes, impacting the quality of Cannabis products and user preferences (Gilbert & DiVerdi, 2018). 1.3.3 Flavonoids Flavonoids are ubiquitous in plants. C-glycoflavones are those most commonly found flavonoids in Cannabis. Early isolations of Cannabis flavonoids revealed the presence of vitexin, orientin, cytisoside, kaempferol, luteolin, apeginen and their glucosides (Clark & 13 Bohm, 1979; Segelman et al 1977). Cannflavins A-C isolated from female Cannabis flowers are methylated isoprenoid flavones, which have been shown to inhibit prostaglandin E2 production, and have anti-leishmanial activity (Barrett et al., 1986; Radwan et al., 2008).  1.3.4 Other phytochemicals Additional secondary metabolites include phytosterols, carotenoids, xanthophylls, stilbenes and alkaloids (Turner et al., 1980; Flores-Sanchez & Verpoorte, 2008). There is limited research into the prevalence, pharmacology and significance of these additional phytochemicals. 1.4 Analytical assays  Analytical methodologies to evaluate phytochemicals in plants can be qualitative to establish the components present or quantitative to determine the concentrations of the specific components of interest. Methods must analyze these components with sufficient accuracy and precision to ensure it is fit-for-purpose which is confirmed using thorough method optimization and validation studies (Mudge et al., 2016a). Method optimization assesses that each parameter of the sample preparation and chromatographic separation are sufficient for the metabolites and materials of interest (Brown & Lister, 2014; Mudge et al., 2016a). Optimization studies are important in natural products due to their complex matrices (Mudge et al., 2016a). Method optimization can include several analyses including factorial designs to select parameters with the largest impact on metabolite extraction and/or individual parameter optimization such as extraction solvent, time, sample mass, etc. (Brown et al., 2010; Mudge et al., 2015a; Mudge et al., 2015b, Mudge et al., 2016b; Mudge et al., 2016c; Mudge & Brown, 2017). There are many standards organizations with validation guidelines, while the one most relevant to plants is AOAC’s single-laboratory validation guidelines for dietary supplements and botanicals (AOAC, 2013; Brown & Lister, 2014; Mudge et al., 2016a). 14 These guidelines ensure validation of sample preparation and analytical procedures as they have significant impacts on the method performance. Method performance in a single laboratory should evaluate several performance characteristics including: selectivity, linearity, repeatability, accuracy, detection and quantitation limits, analyte stability and intermediate precision (Brown et al., 2010; Mudge et al., 2015a; Mudge et al., 2015b, Mudge et al., 2016b; Mudge et al., 2016c; Mudge & Brown, 2017). The following sections summarize the analytical methodologies used to evaluate Cannabis phytochemistry. 1.4.1 Cannabinoids Cannabinoid analysis of THC, CBD and CBN content in Cannabis was originally performed using gas chromatography (GC). Since cannabinoids are naturally present as non-volatile acids in the plants, GC analysis requires either the decarboxylation of cannabinoids or the derivatization of the acids with trimethyl silyl (TMS) or TMS-containing reagents, including N-methyl-N-trimethylsilyltrifluoroacetamide or N,O-bis(trimethylsilyl)trifluoro-acetamide/Trimethylchlorosilane to improve volatility (Fish, 1964; Fetterman et al., 1971; UNODC, 2009). Decarboxylation of dried extracts prior to analysis resulted in losses of cannabinoids due to volatilization of neutral cannabinoids (Kanter et al., 1979; Veress et al., 1990). Decarboxylation of cannabinoids is now performed in the GC injector port, reducing losses during sample preparation, but this limits quantitation to total cannabinoids (Raharjo & Verpoorte, 2004).  Quantitative GC methods for cannabinoids utilize a variety of extraction solvents ranging from ethanol to petroleum ether, while methanol or methanol:chloroform 9:1 % v/v are the most commonly employed (Chan, 2014; Comparini & Centini, 1983; Mehmedic et al., 2010; Elsohly et al., 2000). GC temperature gradients typically range from 170 to 280 °C using 100% dimethylpolysiloxane (DB-1) or 5% diphenyl- and 95% dimethyl-poylsiloxane (DB-5) 15 columns (Chan, 2014; Comparini & Centini, 1983; Mehmedic et al., 2010; Elsohly et al., 2000). Detection is most commonly flame ionization detection (FID) or mass spectrometry (MS).  As GC analysis is limited to neutral cannabinoids rather than as they naturally occur in the plants there was a rapid transition to the development of reversed phase liquid chromatography (LC) methods (Comparini & Centini, 1983; Smith & Vaughan, 1976; Turner & Malhberg, 1982; Lehmann & Brenneisen, 1995; Wheals & Smith, 1975). Recent innovations in analytical instrumentation and columns have enabled the use of smaller columns, novel column chemistries and smaller particle sizes to reduce analysis time and improve peak resolution (Noestheden et al., 2018).  Similar to GC sample preparation there is a lack of consistency between extraction procedures for LC cannabinoid analysis. Original methodologies evaluating qualitative cannabinoid profiles used 80% methanol as the extraction solvent (Wheals & Smith, 1975; Smith 1975). The most commonly used extraction solvent is 9:1 % v/v methanol:chloroform (Mehmedic et al., 2010; De Backer et al., 2009; Swift et al., 2013; Gul et al., 2015; Patel et al., 2017). When originally used, the solvent volume was 1 mL per 100 mg of sample and the justification for the methanol-chloroform mixture was to dissolve the internal standard (di-n-octyl pthalate) and prevent its adsorption to the sample material during extraction (Smith & Vaughan, 1976). Current method preparations consist of 100-500 mg of dried flowers in 3 to 20 mL of 9:1 methanol:chloroform, followed by evaporation of an aliquot and reconstitution in an aqueous methanol solution, or diluted with aqueous methanol prior to injection (Mehmedic e al., 2010; De Backer et al., 2009; Swift et al., 2013; Gul et al., 2015). Cannabinoid quantitation has also been demonstrated with ethanol, methanol and 80:20 acetonitrile:methanol (% v/v) (Giese et al., 2015; Ciolino et al., 2018; Noestheden et al., 2018; Wang et al., 2018). 16 Detection of cannabinoids traditionally employs UV absorbance. The chromophores for cannabinoids have a UV absorbance maximum between 210 to 220 nm, while acidic cannabinoids have an additional UV maximum at 274 nm, which is less sensitive in comparison with 220 nm (Lehmann & Brenniesen, 1995; De Backer et al., 2009). Due to their similarity in chemical structure and hydrophobicity, obtaining adequate resolution between cannabinoids is difficult, therefore UV wavelength selection has been employed to account for these limitations (Upton et al., 2014). Mass spectrometry with positive ionization has been suggested as an alternative to eliminate issues of co-elution of minor components (Stolker et al., 2004; Aizpurua-Olaizola et al., 2014; Wang et al., 2018).  Supercritical fluid chromatography (SFC) has been applied to the quantitation of neutral cannabinoids. Bäckström et al. (1997) performed cannabinoid analysis of ethanol extracts, while Wang et al. (2017) prepared samples with 80:20 acetonitrile:methanol. SFC was shown to be superior to LC and GC due to the improvement in resolution and analysis time.  Nuclear Magnetic Resonance Spectroscopy (NMR) provides a fast screening tool for cannabinoid profiling and quantitation. It provides universal detection of protons in organic molecules and the signal intensity is directly proportional to the molar concentration. Therefore, NMR spectroscopy profiles and quantifies analytes in plant extracts without complex sample preparation or chromatographic separation. For example, Hazekamp et al. (2004) used unique singlet proton signals for five cannabinoids in the range of δ 4.40 to 6.44, by comparing the signal of an internal standard anthracene. NMR analysis took 5 minutes to quantify the major cannabinoids. Semi-quantitative analysis of THC and THCA in water and ethanolic extracts was also performed to evaluate the variations in profiles using more traditional British medicinal preparations (Politi et al., 2008). 17 1.4.2 Terpenes Terpenes are analyzed using gas chromatography coupled with either FID or MS because of their volatility. Originally analysis focused on steam distillation to prepare oils and column chromatography for identification, which evolved to profiling the terpene contents of monoterpenes and sesquiterpenes and determining their relative composition based on peak area of the oils (Nigam et al., 1965; Mediavilla & Steinemann, 1997; Ross & ElSohly, 1996). Other methods used fresh flowers to evaluate the terpenes using organic solvents for direct injection into the instrument (Casano et al., 2011; Hillig, 2004; Romano & Hazekamp, 2013).   Many quantitative methods are generally selective to the known, high abundance terpenes, rather than quantitation of total terpenes. Giese et al. (2015) developed a quantitative method for 17 terpenes in Cannabis flowers using ethanol as an extraction solvent containing nonane as the internal standard with high throughput homogenization prior to GC-FID analysis. The method eliminated losses from sample grinding and solvent evaporation prior to analysis, allowing for direct quantitation of the intact terpenoids. Unfortunately, during the analysis of authentic materials, which were pre-ground, the terpene content decreased over time, resulting in poor inter-day precision, while suitable intra-day precision was obtained (Giese et al., 2015). This highlights the impacts of losses due to volatilization and confirms that ground samples must be analyzed promptly. Previous work by Ross & ElSohly (1996) had shown similar losses of terpenes during storage. Several methods for terpene profiling and/or quantitation are multi-purpose methods to quantify terpenes and total neutral cannabinoids simultaneously (Elzinga et al., 2015; Hazekamp & Fischedick, 2012; Fischedick et al., 2010).  18 1.4.3 Other phytochemicals Profiling and quantitation methods for other classes of compounds detected in Cannabis are lacking. In 1979, the variation of nine flavonoids in Cannabis was determined by evaluating their presence/absence using thin-layer chromatography (TLC) (Clark & Bohm, 1979). An HPLC method was used to quantify four flavonoids in fiber-type Cannabis using UV and MS detection (Vanhoenacker et al., 2002). Isahq et al. (2015) evaluated the presence of alkaloids, saponins, tanins, phenols and flavonoids in Cannabis indica leaves, seeds and stems, although method specifics were not provided. 1.5 Metabolomic evaluations Metabolomics is a technique to understand the phytochemical variation in plants based on genetic, environmental or developmental changes (Hall, 2011; Turi et al., 2015). Functional genomics is the backbone to metabolomic research as the gene expression impacts the phenotype, which can be defined by the chemical characteristics of the plants, among others (Feihn, 2002; Nicholson et al., 2002). Metabolites can either be defined as primary or secondary, where primary are necessary for basic functions of the cells of plants and secondary metabolites are not necessary for function but play important roles in environmental interaction and survival. Metabolomics studies the small, secondary metabolites produced by the organisms (Hall, 2011). The goal is to view the plant metabolites at a specific time, in a specific tissues, cells or organelles under specific conditions.  Plant metabolomics has commonly been classified as either targeted analysis, metabolomic profiling, or metabolomic fingerprinting (Turi et al., 2015). Targeted analysis is traditional separation and quantitation of specific, known phytochemical classes in plants. These methods are highly selective and do not provide an entire picture of the plant metabolome. Untargeted metabolomics encompasses metabolomic profiling and 19 fingerprinting. Metabolomic fingerprinting evaluates the patterns generated by a specific sample, using a particular analytical technique. Comparisons are made to control samples, and quantitative data is generally not acquired. Metabolomic profiling delves deeper by elucidating and potentially quantifying the structures of metabolites identified as unique between the different plants (Hall, 2011). In these two cases, thousands of metabolites may be detected and used to define the variation of different plants. The benefits of untargeted approaches is that there is a higher probability of identifying the relative changes in metabolites due to the wide variety of chemical classes that are measured (Choi & Verpoorte, 2014).  Metabolomics experiments follow several basic procedures: sample harvesting and preparation, extraction, analysis, data interpretation and metabolite identification. Variation of any of these parameters will highly impact the resulting metabolome. Data sets generated from metabolomics analysis are substantial and complex, therefore multivariate analysis is employed. Principal component analysis (PCA) is an unsupervised model frequently used as a way to view the differences in the data but can have limitations when variation within groups is high (Sumner et al., 2003; Worley & Powers, 2013). Supervised models such as Partial Least Squares Discriminant Analysis (PLS-DA) considers the pre-defined groupings of the samples to reduce the data (Berreuta et al., 2007). Both techniques have the ability to discriminate samples or treatments and identify unique metabolites.  1.5.1 Forensic applications Forensic determination of Cannabis origins was established in the US to manage drug intelligence and monitor drug trafficking. The University of Mississippi completed a pilot study to determine plant origin with GC-FID and LC-UV profiling using base peak ratios of over 100 different peaks in the chromatograms. They identified unique or characteristic peaks for each of the five regions studied (Brenneisen & ElSohly, 1988).  20  The second phase of this study evaluated GC-MS profiles of Cannabis extracts from known geographical origins, coupled with multivariate analysis. 175 compounds were used to characterize the materials, of which 46 were known constituents. The models identified all but one sample correctly when determining domestic versus foreign origin, where a Hawaiian sample was identified as foreign. When establishing the origin of foreign materials, more than 90% of materials were correctly identified (ElSohly et al., 2007). The data assumed that confiscated materials were grown in the same region. Using this same profiling, it was possible to identify indoor versus outdoor produced marijuana, maturity and sex of the plant (ElSohly et al., 2007).   Aside from the research performed in the United States, cannabinoid profiles determined using head space solid phase microextraction GC-MS analysis confirmed regional differences of Cannabis flowers within Switzerland and profiling of hashish samples were used to determine if samples were from the same batch in Switzerland and Malaysia (Cadola et al., 2013; Chan et al., 2014; Ilias et al., 2005). 1.5.2 Chemotaxonomy applications As previously mentioned, the classification of Cannabis is still debated between taxonomists, while the majority of published research considers all Cannabis as Cannabis sativa L. to remain consistent with regulatory classifications of the plant (Small, 2015). Original phenotypes established by Small & Beckstead (1973) identified three classes: phenotype I containing high THC:CBD ratios of >1.0, phenotype II containing midrange levels of THC:CBD 0.5-2.0, and phenotype III is the fiber types with THC:CBD ratios <1.0.   These phenotypes are regularly referenced in scientific literature to classify Cannabis products, but there is an increased interest in using analytical methods to further differentiate Cannabis species based on their origin, chemical composition and morphological 21 characteristics. Hillig (2005) evaluated 157 Cannabis materials of known geographical origin for allozyme variations at 17 gene loci and classified three species of Cannabis (C. sativa, C. indica and C. ruderalis), with 7 putative taxa. Following genetic evaluations, the same accessions were assessed for cannabinoid contents using GC-FID analysis of neutral cannabinoids. The THC:CBD ratios were classified into three distinct chemotypes similar to that previously defined (Hillig & Malhberg, 2004). The C. indica biotypes were associated with high levels of THC, although some fiber biotypes also had higher levels of THC, while C. sativa biotypes had low THC contents. Due to the identification of all three chemotypes within the biotypes and species, it was concluded that cannabinoid contents did not have a significant impact on chemotaxonomic classification, and only supports the two-species concept rather than the genetic variation three-species concept previously proposed (Hillig & Malhberg, 2004).  Terpene profiles were also assessed to determine their potential to establish taxonomic classification of Cannabis. Using GC-FID, a total of 48 known and unidentified peaks corresponding to mono- and sesquiterpenes were used with PCA analysis (Hillig, 2004). 91% correct assignment based on the two-species concept, therefore eliminating the proposed C. ruderalis species using chemotaxonomic classification. 1.5.3 Strain differentiation Terpene and cannabinoid profiles are the most common secondary metabolites for evaluating variation in different strains of Cannabis. Terpene profiles were used to evaluate 16 strains of Cannabis of which half were considered ‘mostly sativa’ and the other ‘mostly indica’. Variation between strains were significant as β-myrcene with α-pinene or limonene were most abundant in ‘mostly indica’ strains, while ‘mostly sativa’ strains were more unpredictable with α-pinene or α-terpinolene most abundant or β-myrcene with α-terpinolene or trans-β-ocimene (Casano et al., 2011).  22 In the Netherlands, one ‘sativa’ designated strain and one ‘indica’ designated strain with similar THC contents were purchased from multiple coffee shop locations and were compared with pharmaceutical grade Cannabis (Hazekamp & Fischedick, 2012). In total, 28 terpenes and cannabinoids were quantified in the materials and used for comparisons. The two strains were chemically distinct based on their terpene profiles where the ‘sativa’ strain was characterized by terpinolene, α-guaiene and γ-selinene, while the ‘indica’ strain was high in α-pinene, and β-myrcene (Hazekamp & Fischedick, 2012). Variations greater than 25% were found in the coffee house samples, while the pharmaceutical grade Cannabis had no more than 11% variability. Expansion of this work was undertaken to evaluate cannabinoids and terpenes profiles of 35 different strains, composed of ‘sativa’, ‘indica’ and ‘hybrid’ as specified by the Leafy Cannabis strain database.  A total of 31 compounds were evaluated in this study. With an increased number of replicates, strains and phytochemicals it was not possible to differentiate the strains using principle component analysis (Elzinga et al., 2015). Alternatively, they observed a “chemical continuum” across the variety of strains. The authors noted that the natural variance within strains was high, in addition to losses during storage (Elzinga et al., 2015).  An additional targeted metabolomics approach compared the terpene and cannabinoid contents of 11 strains grown under identical conditions and found a positive correlation between the total terpene and total cannabinoid content (Fischedick et al., 2010). This quantitative metabolomics approach, which detected 36 terpenes and cannabinoids, was able to distinguish each variety of Cannabis using PCA. Another study evaluated 460 different Cannabis accessions for 44 terpenes and cannabinoids to establish chemical varieties (chemovars) where hydroxylated terpenes were more prevalent in ‘indica’ classified strains (Hazekamp et al., 2016). Another classification model was developed with 32 strains collected from 2 Canadian licensed producers by evaluating 24 cannabinoids and terpenes. Their strain 23 selection appeared limited with THC contents ranging from 0.24 to 7.08 % but identified 4 chemotaxonomic clusters through multivariate analysis (Jin et al., 2017).  The studies described previously are examples of targeted metabolomics, involving the analysis and profiling of known components to classify the different strains. The first untargeted metabolomics study on Cannabis was performed by Choi et al. (2004) using NMR. The spectra generated for crude extracts of 12 cultivars of Cannabis were used to differentiate medicinal or “drug-type” cultivars of Cannabis sold in the Netherlands. A two-phase extraction solvent composed of 50% aqueous methanol and chloroform was prepared and the organic and aqueous phases were analyzed separately. THCA was the predominant component in the chloroform extracts, while differentiation was established based on THCA and CBDA contents in the different cultivars using PCA. The aqueous extracts contained a significant amount of amino acids and sugars which were used for discriminating species (Choi et al., 2004).  Untargeted metabolomic studies using several chemometric models to classify and authenticate 25 individual Cannabis strains have been highlighted with NMR and high-resolution mass spectrometry (Wang et al., 2017; Wang et al., 2018). The classification methods implemented to evaluate classification performance included fussy-rule building expert system, linear discriminant analysis, super partial least squares discriminant analysis, support vector machines and support vector machines classification trees. The resulting data presented the robustness and performance of models to authenticate each strain as individual classes, indicating that each strain can be classified individually and is chemically distinct in some capacity (Wang et al., 2017; Wang et al., 2018). 24 1.6 Study rationale and objectives With increased acceptance, use and legal production of Cannabis for medical conditions, there is a strong need to evaluate and understand phytochemical diversity of strains. The overall objective of this thesis is to establish new techniques to correlate the Cannabis metabolome and identify underlying relationships. These relationships will contribute to understanding domestication syndrome and pharmacological variance.  Cannabis breeding practices to develop new ‘strains’ have lacked rigorous plant characterizations, relying primarily on phenotypic and organoleptic evaluations. Additionally, current literature on Cannabis phytochemistry focuses on only high abundance metabolites, where minor and untargeted metabolomics have not been thoroughly investigated. The objectives of this study were: 1. To develop, optimize and validate a quantitative analytical method for accurate determination of cannabinoids in Cannabis flowers. 2. To use the validated quantitative method to evaluate the cannabinoid diversity across several Cannabis strains available in the Canadian marketplace and to use this data to identify strain clusters, metabolite relationships and evaluate the impacts of domestication and breeding on cannabinoid profiles. 3. To develop a methodology to evaluate the terpene profiles of the Cannabis strains and identify relationships between terpenes and strain clusters. This approach was used to evaluate the pharmacological significance of different terpene clusters and identify the impacts of domestication and breeding on terpene profiles.  4. To use chemometric tools to evaluate the underlying chemical diversity and explore in-depth relationships between cannabinoids, monoterpenes and sesquiterpenes. Identify any biochemical breaks in the cannabinoid and terpenes profiles. 25 5. To isolate and elucidate the structures of unknown compounds highlighted in the metabolite relationships and to determine potential connection to breeding and domestication syndrome.   26 Chapter 2: Single laboratory validation of an HPLC-UV method to quantify cannabinoids in Cannabis  2.1 Synopsis Phytochemical characterization for quantifying marker or pharmacologically significant metabolites requires rigorously validated methods to ensure they are fit-for-purpose. For Cannabis, the most prevalent metabolites of interest are the cannabinoids, which are the prenylated polyketides found in the glandular trichomes predominantly on the female inflorescences. Regulatory requirements within Canada require licensed producers to quantify the contents of CBDA, CBD, Δ9-THCA and Δ9-THC using validated methods (Health Canada, 2018). There are many methods publicly available for the quantitation of cannabinoids which lack sufficient optimization and validation to establish the method performance and reliability. The extraction parameters including solvent composition, mass-to-solvent volume ratio, and extraction technique all vary significantly between published methods as well as chromatographic separation conditions. The most commonly employed extraction solvent for cannabinoids analysis is 9:1 methanol:chloroform (% v/v), with some exceptions (Lehman & Brenneisen, 1996; Swift et al, 2013; De Backer et al., 2009; Mehmedic et al., 2009; Smith & Vaughan, 1976). The original purpose was to dissolve the internal standard di-n-octyl phthalate, which could be determined with GC analysis, and is no longer necessary with improved instrument performance and commercially-available reference standards (Smith & Vaughan, 1976). There has been an increased desire to remove chlorinated solvents from laboratory analysis whenever possible due to its possible toxicity, cost of disposal and hazards in transport and storage (Watts et al., 2004; Alfonsi et al., 2008; CDC, 2016). Long-term exposure to chloroform is associated with liver and kidney damage when the occupational exposure limit is 2 ppm in the air (Watts et al., 2004; CDC, 2016). Removal of chloroform from 27 the extraction solvent will improve laboratory safety, reduce reagent and disposal costs and improve the environmental impact of the method.  The HPLC method described quantifies eight cannabinoids and was adapted from previously published methods (CBDA, Δ9-THCA, CBD, Δ9-THC, Δ8-THC, THCV, CBG, CBN and CBC). The simplified, green chemistry method was subjected to a single-laboratory validation after implementing a statistically guided method development protocol to ensure optimization of the sample preparation procedures. Nine authentic Cannabis flower materials were used in the validation to evaluate the method performance. A method validation extension was also undertaken to evaluate the linearity and repeatability of four additional cannabinoids (CBDVA, CBDV, CBGA and CBL). 2.2 Experimental 2.2.1 Reagents Methanol and acetonitrile were HPLC grade and purchased from VWR International (Mississauga, ON). ACS grade chloroform was also purchased from VWR International. Water was purified to 18 MΩ with a Barnstead Smart2Pure nanopure system (Thermo Scientific, Waltham, MA). Ammonium formate and formic acid were >98% pure and HPLC grade and purchased from Sigma Aldrich (Oakville, ON) and Fisher Scientific (Ottawa, ON), respectively. 2.2.2 Calibration standards Certified reference materials (CRMs) were purchased from Cerilliant Corp. (Round Rock, TX) for thirteen cannabinoids: Δ9-THCA, Δ9-THC, CBDA, CBD, cannabigerol (CBG), cannabichromene (CBC), tetrahydrocannabivarin (THCV), cannabinol (CBN), Δ8-THC, cannabidivarinic acid (CBDVA), cannabidivarin (CBDV), cannabigerolic acid (CBGA) and cannabicyclol (CBL). The individual cannabinoids were provided in solution at 1.0 mg/mL 28 concentration certified by the supplier. The acidic cannabinoids were provided in acetonitrile and neutral cannabinoids in methanol. Fresh ampules were used for the validation study to ensure accurate quantitation. Due to the release of additional cannabinoids by the supplier, the original validation was performed using the first 9 cannabinoids. A method extension was performed to evaluate repeatability with the additional four cannabinoids at a later date.  2.2.3 Test materials Dried medical Cannabis samples were purchased from several licensed producers within Canada. Nine products were selected with a variety of cannabinoid concentrations ranging from 0.2 to 17% total THC and 0.3 to 9 % total CBD according to the supplier labels. All analyses were performed in a laboratory holding a controlled substance license permitting Cannabis analysis. Seven additional Cannabis samples were used to evaluate the repeatability of the four additional method-extension cannabinoids. 2.2.4 HPLC analysis An Agilent 1200 RRLC system equipped with a temperature-controlled autosampler, binary pump and diode array detector was used to separate the cannabinoids (Mississauga, ON). The separation was achieved on a Kinetex® C18, 100 x 3.0 mm, 1.7 µm column (Phenomenex, Torrance, CA). The mobile phase was composed of (A) 10 mM ammonium formate, pH 3.6 (B) acetonitrile filtered to 0.2 µm. Gradient conditions with a flow rate of 0.6 mL/min was achieved according to the following: 0–8 min, 52–66%B; 8–8.5 min, 66–70%B; 8.5–13 min, 70–80%B; 13–15 min, 80%B. A 7-minute column equilibration was performed after each run. The injection volume was 5 µL, the detection was at 220 nm, the column temperature was 25 °C and the autosampler was maintained at 4 °C. 29 2.2.5 Method optimization Ground material from two Cannabis test samples: a mid-level THC/CBD strain and a high-THC strain were used for the method optimization. 2.2.5.1 Fractional factorial The partial factorial for method optimization and data analysis was designed using Minitab 16 (Minitab, State College, PA). Individual cannabinoids were quantified as weight percentage in Cannabis flowers. Additional optimization parameters were evaluated based on the performance within the factorial design. Microsoft Excel was used to calculate cannabinoid contents and statistical analysis of the validation data. 2.2.5.2 Analyte stability Mixed calibration standards were stored at -20 °C, 4 °C and 22 °C in the dark and tested at regular intervals to assess cannabinoid stability in solutions. Sample extracts were stored at 4 °C and 22 °C in light and dark conditions. A sample with greater than 5% variation from time zero was considered unstable. 2.2.6 Preparation of test materials A minimum of 5 grams of dried flowers were ground together for each test sample to ensure homogeneity. Ground flowers were extracted by weighing 200.0 mg into a 50 mL amber centrifuge tube and adding 25.00 mL of 80% aqueous methanol. The solution was vortexed for 30 seconds and extracted for 15 minutes in a sonicating bath, with vortexing every 5 minutes. Extracts were filtered with 0.22 µm PTFE filters and diluted 1:2, 1:5 or 1:10 using the extraction solvent into an amber HPLC vial and stored at 4 °C prior to analysis. 30 2.2.7 Single-laboratory validation parameters The optimized method was subjected to a single-laboratory validation according to AOAC International guidelines for dietary supplements (AOAC, 2013). Δ8-THC, CBDV and CBL were not observed in any of the samples and were not evaluated for repeatability. 2.2.7.1 Selectivity Selectivity was demonstrated by injecting the reference materials and flower extracts to evaluate the resolution between closely eluting peaks and potential interferences at 220 nm. Resolution greater than 1.5 is deemed acceptable by AOAC guidelines for complex mixtures (AOAC, 2013). Peak purity was verified for all cannabinoids of interest. 2.2.7.2 Linearity Individual cannabinoid CRMs were used to prepare seven-point standard calibration curves for the original eight cannabinoids in concentrations ranging from 0.5 to 250 µg/mL. Dilutions of the CRMs were performed using the extraction solvent (80% methanol). Concentration ranges were modified for each cannabinoid as summarized in Table 2.1. The calibration curves were plotted and the slope and y-intercept for each cannabinoid were used for linear regression analysis. Calibration curves were visually inspected, and correlation coefficients were determined. An r2 above 0.995 was deemed suitable for quantitation. Mixed standards were stored at 4 °C and were stable for up to 3 days.   31 Table 2.1 Concentrations of the cannabinoids used in the mixed calibration standards for the seven-point calibration curves. Lin = calibration standard label  Concentration (g mL-1) Cannabinoid Lin 1 Lin 2 Lin 3 Lin 4 Lin 5 Lin 6 Lin 7 CBDA 250 200 100 50 25 10 5 THCV 25 20 10 5 2.5 1 0.5 CBD 50 40 20 10 5 2 1 CBG 25 20 10 5 2.5 1 0.5 CBN 25 20 10 5 2.5 1 0.5 THCA 250 200 100 50 25 10 5 THC 50 40 20 10 5 2 1 CBC 25 20 10 5 2.5 1 0.5  2.2.7.3 Repeatability and intermediate precision Quadruplicate samples of each test material were prepared on a single day to evaluate the repeatability as relative standard deviation (% RSD) for the original 8 individual cannabinoids. Intermediate precision was determined by repeating the repeatability studies on three separate days. The within-day, between-day and total standard deviations were calculated for each cannabinoid in each test material. HorRat values were calculated to assess the intermediate precision of the method (Horwitz, 1982). Triplicate samples of seven test materials were prepared on a single day to evaluate the repeatability as relative standard deviation (% RSD) for the additional 2 method extension cannabinoids present in the samples (CBGA and CBDVA).  2.2.7.4 Recovery Recovery was determined at three concentration levels for the major cannabinoids: CBDA, CBD, THCA and THC. Ground stinging nettle, used as a matrix blank was spiked with individual cannabinoids and prepared according to the samples preparation protocol. The quantified content of the individual cannabinoids was compared with the known concentration to calculate the percent recovery.  32 2.2.7.5 Limits of detection and quantitation The limits of detection and quantitation were determined using the US Environmental Protection Agency (EPA) method detection limit (MDL) protocol (EPA, 2002). The MDL is defined as the minimum concentration of substance that can be measured and reported with 99% confidence that the analyte concentration is greater than zero. Extract solutions containing low concentrations of the cannabinoids were used to evaluate the method limits. Seven replicates were injected and the calculation for MDL was determined as the standard deviation of the calculated concentration between the seven replicates multiplied by the t-statistic at 99% confidence interval. LOQ was determined as 10 times the standard deviation for the replicates to determine the MDL. 2.3 Results 2.3.1 Method optimization A statistically-guided optimization plan was developed using a partial factorial design to determine the impact of four parameters used in cannabinoid extractions from dried flowers. The parameters included: extraction time, extraction technique, extraction composition and solvent volume-to-mass ratio. A large variance in the two levels would indicate that this parameter significantly impacts the extraction and requires further evaluation (Mudge et al., 2016b). The runs performed for the factorial design study are summarized in Table 2.2.   33 Table 2.2 Runs prepared for the partial factorial design to optimize the sample preparation of cannabinoid extractions. Run No. Extraction Time (min) Extraction Technique Solvent Volume (mL) Solvent Composition 1 15 Shaking 25  MeOH:CHCl3* 2 60 Shaking 25  80% MeOH 3 15 Sonication 25  80% MeOH 4 60 Sonication 25  MeOH:CHCl3 5 15 Shaking 10  MeOH:CHCl3 6 60 Shaking 10 80% MeOH 7 15 Sonication 10  80% MeOH 8 60 Sonication 10  MeOH:CHCl3    * 9:1 v/v 2.3.1.1 Extraction of tissues Two levels were selected for each factor. The statistical analysis of these data indicated that solvent volume was the most significant factor, followed by solvent composition (Figure 2.1). Extraction technique and time did not affect the extraction of cannabinoids.   Figure 2.1 Variation of the total cannabinoid content determined using the partial factorial design to optimize the sample preparation.  34 A further in-depth evaluation of the extraction parameters was undertaken. The solvent volume-to-mass ratio and solvent composition using 25 mL extraction solvent with 100 and 200 mg of sample confirmed that 200 mg was equivalent to 100 mg, without saturation of the extraction solvent (Figure 2.2). The improved precision observed with 200 mg of sample, indicated that 200 mg may be better suited to reduce inhomogeneity issues with Cannabis flowers. The composition of the extraction solvent did not impact the content of the major cannabinoids. An overlay of the chromatographic separation of the two extraction solvents also confirms that the cannabinoid composition is identical between the two extraction solvents (Figure 2.3). In order to evaluate the effectiveness of a single extraction, a comparison of a single extraction with 25 mL was compared with 2 x 10 mL, which were pooled into a 25 mL volumetric flask. The data were not significantly different between the two extractions, therefore a single extraction was sufficient with the solvent volume-to-mass ratio selected (Figure 2.4).  Figure 2.2 Cannabinoid content of CBDA, CBD, THCA and THC evaluated using two different extraction solvents, 9:1 v/v methanol:chloroform and 80% methanol, and two different sample masses, 100 and 200 mg. Preparations of n=3 replicates, error bars represent standard error of the mean (SEM).   35  Figure 2.3 Overlay of the chromatographic separation of cannabinoids extracted with 9:1 methanol:chloroform (black) and 80% methanol (red).   Figure 2.4 Comparison of a single extraction versus a 2 times extraction to confirm extraction effectiveness. Error bars represent SEM, n=3.   Although extraction time was not indicated as a significant factor, it was optimized to increase sample throughput. The factorial design showed slightly lower total cannabinoids at 60 minutes in comparison to 15 minutes, potentially indicating some degradation during the longer extractions. Three time points were assessed: 15, 30 and 60 minutes. The level of cannabinoids was not significantly different between the time points (Figure 2.5). It was then verified if extraction time could be reduced by evaluating 5, 10 and 15 minutes in comparison 0123456CBDA CBD THCA THCConcentration (% w/w)Single ExtractionDouble Extraction36 to 15 minutes with vortexing every 5 minutes. Again, no significant difference was observed between all four extraction times, while 15 minutes with vortexing was significantly higher than 5 minutes (Figure 2.6). These findings were used to formulate the sample preparation protocol using 200 mg of dried flowers with 25 mL of 80% aqueous methanol for 15 minutes by sonication with vortexing every 5 minutes.   Figure 2.5 Cannabinoid content of CBDA, CBD, THCA and THC after sonication for 15, 30 and 60 minutes. Error bars represent SEM, n=3.  37  Figure 2.6 Cannabinoid content of CBDA, CBD, THCA and THC after sonication for 5, 10, 15 minutes and 15 minutes with vortexing every 5 minutes. Error bars represent SEM, n=3.  2.3.1.2 Chromatographic optimization Several HPLC columns, phases, dimensions, mobile phases and gradients were compared in preliminary experiments to determine a method for baseline separation of as many cannabinoids as possible while maintaining a short separation time. The optimal column used for separating cannabinoids was a Phenomenex Kinetex core shell C18 column with a sub 2.0-micron particle size. The typical pump pressure required was approximately 400 bar, therefore a 600-bar instrument was necessary for this analysis. Mobile phase pH appeared to be a significant factor in cannabinoid separation; as pH decreases the retention of acidic cannabinoids increases, while at higher pH the retention decreases with poor peak shape. The buffer at pH 3.6 was found to improve baseline separation of THCA from other cannabinoids, while maintaining adequate peak shape. The final analysis was a 15-minute run with gradient elution using 10 mM ammonium formate at pH 3.6 and acetonitrile as mobile phases. The final chromatogram is shown in Figure 2.7a. Resolution of major cannabinoids 38 7were greater than 2.0, while minor cannabinoids were greater than 1.0 using the mixed calibration standards. Sample extracts were used to confirm the resolution and peak purity during the validation study (Figure 2.7b).  Figure 2.7 Chromatographic separation of cannabinoids using a Kinetex C18 column (100 x 2.1 mm, 1.8 µm) with gradient elution at 220 nm. (A) mixture of cannabinoids standards (B) authentic Cannabis extract.  39 2.3.1.3 Analyte stability The stability of cannabinoids was assessed to determine whether losses occur prior to analysis that may significantly impact quantitation. Mixed calibration standards prepared in 80% aqueous methanol were stored at -20 °C, 4 °C and 22 °C in the dark. Variations of less than 5 % compared to the time zero data were considered acceptable. The losses are summarized in Table 2.3. Losses greater than 5% from time zero found after 30 h at room temperature with CBDA/THCA contents decreasing by 6.3 and 9.6 % after 48 hours, while mixed solutions stored at -20 °C resulted in losses of 8.1 and 10.6 %, respectively. The standards were stable at 4 °C for the duration of the 72 hour study. Table 2.3 Cannabinoid stability of mixed standard solution prepared in 80% methanol stored at three different temperatures in the dark. Parentheses indicate increases in concentration (n=3). Time (hrs) Room Temperature -20 °C 4 °C CBDA CBD THCA THC CBDA CBD THCA THC CBDA CBD THCA THC 6 1.1 0.6 1.6 0.9         24 4.4 0.0 4.6 0.3 1.7 1.4 1.6 1.6 1.8 (0.4) 2.0 (0.1) 30 6.3 3.6 9.6 1.4         42         1.9 (0.8) 3.3 (2.2) 48 2.5 (4.2) 4.8 (12.7) 8.1 6.2 10.6 4.5 2.4 (1.8) 1.7 (1.7) 72     14.4 13.8 14.5 8.3 1.0 (2.5) 2.0 (2.5) No entry indicates timepoints that were not evaluated. Sample extracts were also prepared in 80% methanol and stored at 22 °C in light and dark conditions. Under these conditions, THCA and CBDA were considered unstable after 24 hours with reductions of 6.7 % for both, resulting in 8-10% increases in the neutral forms of these cannabinoids (Table 2.4). Reductions of 11 to 23 % of acidic cannabinoids occurred under light conditions with 24 hours. On the basis of these findings an additional study was performed to evaluate extract stability in 80% aqueous methanol and 9:1 methanol:chloroform stored in the dark at 22 °C and 4°C (Table 2.5 and Table 2.6). The 9:1 methanol:chloroform extracts were found to be stable at 22 °C and 4 °C for the duration of the study, 36 and 48 hours, respectively, while 80% methanol were stable at 4°C for 48 hours. The extractions in 80% methanol were not stable at room temperature for 24 hours, similar to the previous study. 40 The elimination of chloroform from the extraction solvent improves analyst safety, reduces solvent costs and does not impact the quantitative results; therefore 80% methanol is a viable alternative extraction solvent for quantitative analysis with the use of a dark, temperature-controlled autosampler. Table 2.4 Cannabinoid stability of authentic Cannabis extracts in 80% methanol at room temperature after 24 hours in dark and light conditions. Parentheses indicate increases in concentration (n=3). Conditions Sample 1 Sample 2 CBDA CBD THCA THC THCA THC 24 hr dark 3.3 (9.1) 3.2 (10.1) 2.3 (16.7) 24 hr light 18.1 (2.5) 22.6 0.8 11.5 0.3  Table 2.5 Cannabinoid stability in authentic Cannabis extracts in 80% methanol at 4 and 22 °C for 48 hours. Parentheses indicate increases in concentration (n=3). Time (hr) 80% Methanol 4 °C 80% Methanol 22 °C CBDA CBD THCA THC CBDA CBD THCA THC 24 1.2 (4.8) 0.9 1.6 6.7 (2.2) 6.7 (8.1) 48 1.5 (5.6) 0.9 0.3 5.4 (8.3) 5.2 (13.3) % change total CBD/THC 2.0 0.9 3.8 3.0  Table 2.6 Cannabinoid stability in authentic Cannabis extracts in 9:1 v/v methanol:chloroform stored at 4 and 22  °C for 48 hours. Parentheses indicate increases in concentration (n=3). Time (hr) 9:1 Methanol: Chloroform 4 °C 9:1 Methanol: Chloroform 22 °C CBDA CBD THCA THC CBDA CBD THCA THC 24 0.1 (2.1) 0.1 (2.9) (0.2) 6.3 (0.4) 0.0 48 0.1 (0.1) 0.2 (4.6) (0.2) 5.9 (0.5) 0.3 % change total CBD/THC 0.2 0.8 0.5 0.4  2.3.2 Method validation 2.3.2.1 Selectivity The chromatographic profiles of Cannabis extracts at 220 nm were used to evaluate peak resolution, which ranged from 1.64 to greater than 2.0 as specified in Table 2.7; in 41 accordance with AOAC guidelines for dietary supplements and botanicals, a resolution greater than 1.5 is sufficient for quantitation given the complexity of natural products (AOAC, 2013). Sample extracts were used to evaluate peak purity and confirm resolution with minor peaks. Peak purity was greater than 98% for all cannabinoids evaluated in this assay. Table 2.7 Resolution and peak purity for each cannabinoid used to determine method selectivity. Cannabinoid Resolution Peak Purity (%) CBDA 6.34 99.8 THCV 1.75 99.9 CBD 1.64 98.8 CBG 1.85 99.9  CBN 3.18 99.9 THCA 1.95 99.5 THC 2.89 99.9 CBC 2.29 99.8  2.3.2.2 Linearity The seven-point calibration curves used on each day of the validation were linear on visual inspection. The correlation coefficients (r2) for each cannabinoid was greater than 0.995 for all calibration curves on each day of the analysis, as summarized in Table 2.8. The plots of residuals were random, confirming that linear functions were suitable for cannabinoid concentrations up to 250 µg/mL. Concentrations of CBDA in the high-THC products CAN004 and CAN007 were lower than the lowest concentration standards using in the validation study. In this case, the materials are outside the working range of the method.  Table 2.8 Linear regression data for each cannabinoid. Cannabinoid Typical Calibration Equation Average correlation coefficients (r2) CBDA y = 11.276 x - 4.6832 0.9990 THCV y = 10.021 x – 0.6974 0.9993 CBD y = 9.0905 x + 5.537 0.9982 CBG y = 9.8255 x – 0.4786 0.9995 CBN y = 15.162 x + 6.9901 0.9992 THCA y = 11.238 x + 7.6676 0.9992 THC y = 9.3853 x + 4.8576 0.9983 CBC y = 9.1987 x -0.6967 0.9991  42 2.3.2.3 Repeatability Repeatability was assessed by quantifying the eight cannabinoids for which standards were available. The quantitative data from the four replicate samples on day 1 were used to determine method repeatability. All precision measurements used authentic Cannabis materials with a range of cannabinoid concentrations. The repeatability data are summarized in Table 2.9. Repeatability RSDs ranged from 0.78 to 10.08 %. For materials with higher than 0.5% w/w cannabinoid content, the % RSDs ranged from 0.78 to 7.64 %, with only two materials having RSDs greater than 5%. Given that the precision for seven of the nine materials evaluated was less than 5%, it is possible that the poorer precision for the other two materials was due to inherent variability of the cannabinoids within these two test samples, rather than an indication of method performance. These two strains were purchased from the same producer which may have a higher inherent variability compared to other products used in the study. The RSDs over 5% for all other materials were observed in low level cannabinoids, for which small variations in the quantitative data will have more impact on the % RSDs. These values are within acceptable validation limits based on their concentrations (AOAC, 2013). Table 2.9 Repeatability and intermediate precision for cannabinoid quantitation in Cannabis dried flowers. Sample ID Cannabinoid Concentration (% w/w) Repeatability (% RSD) Intermediate Precision (% RSD) HorRat Cannabis flowers  ID # CAN001  CBDA 3.61 2.74 4.26 1.3 CBD 0.22 3.82 6.01 1.2 CBG 0.07 6.19 7.13 1.2 CBN 0.02 8.45 9.92 1.4 THCA 7.81 3.40 2.85 1.3 THC 0.67 3.32 4.44 1.1 CBC 0.03 6.12 11.41 1.7 Cannabis flowers  ID # CAN002  CBDA 3.70 2.86 3.78 1.2 CBD 0.49 2.68 2.96 0.7 CBG 0.02 10.08 10.08 1.4 THCA 0.11 3.29 3.17 0.6 THC 0.03 3.56 11.61 1.7 CBC 0.04 2.17 2.97 0.5 43 Sample ID Cannabinoid Concentration (% w/w) Repeatability (% RSD) Intermediate Precision (% RSD) HorRat Cannabis flowers  ID # CAN003  CBDA 5.97 2.35 3.32 1.0 CBD 0.25 2.05 4.75 1.0 CBG 0.10 1.51 3.96 0.7 CBN 0.02 3.90 7.77 1.1 THCA 10.3 1.80 3.65 1.3 THC 0.84 1.75 3.25 0.8 CBC 0.03 4.49 7.40 1.1 Cannabis flowers  ID # CAN004  CBDA 0.04 2.97 11.13 1.7 CBG 0.12 3.78 4.02 0.7 CBN 0.05 3.17 8.06 1.4 THCA 13.9 2.62 3.50 1.3 THC 1.74 2.39 3.21 0.9 CBC 0.06 3.63 6.37 1.0 Cannabis flowers  ID # CAN005  CBDA 7.81 1.76 2.69 0.9 CBD 1.43 1.51 2.08 0.6 CBG 0.06 1.97 4.92 0.8 CBN 0.06 4.29 3.96 0.7 THCA 2.90 1.84 2.96 0.9 THC 0.95 2.03 3.20 0.8 CBC 0.09 3.57 3.53 0.6 Cannabis flowers  ID # CAN006  CBDA 6.27 1.61 4.09 1.4 CBD 0.70 1.08 3.48 0.8 CBG 0.11 1.06 5.78 1.0 CBN 0.04 4.94 4.34 0.7 THCA 8.67 0.78 5.64 2.0 THC 1.06 1.35 3.75 1.0 CBC 0.05 3.50 3.73 0.6 Cannabis flowers  ID # CAN007  CBDA 0.04 4.29 11.67 1.8 THCV 0.03 4.26 5.24 0.8 CBG 0.24 4.77 5.06 1.0 CBN 0.15 2.77 3.43 0.7 THCA 14.9 2.72 4.66 1.8 THC 3.31 3.64 4.01 1.2 CBC 0.05 7.72 7.80 1.3 Cannabis flowers  ID # CAN008  CBDA 8.14 5.84 5.16 1.8 CBD 0.56 4.47 5.28 1.2 CBG 0.07 6.85 7.00 1.2 CBN 0.03 7.64 6.65 1.0 THCA 5.39 6.18 5.36 1.7 THC 0.88 4.53 5.57 1.4 CBC 0.05 4.20 5.44 0.9 Cannabis flowers  ID # CAN009  CBDA 7.95 7.24 5.94 2.0 CBD 0.36 4.74 5.11 1.1 CBG 0.06 4.51 6.28 1.0 CBN 0.06 3.29 10.87 1.3 THCA 2.05 7.64 6.65 1.9 THC 0.27 4.43 3.77 0.8 CBC 0.03 2.28 4.47 0.7 44  Due to the release of additional cannabinoid references standards after the completion of the original validation, repeatability of two additional cannabinoids (CBGA, and CBDVA) were evaluated on seven samples prepared in triplicate. CBGA had % RSD values from 1.3 to 5.9 %, and CBDVA ranged from 1.6 to 5.1 % as summarized in Table 2.10.  Table 2.10 Repeatability (% RSD) for the additional cannabinoids detected in Cannabis flowers which were assessed after the original validation was completed. Sample ID Cannabinoid Concentration (% w/w) Repeatability (% RSD) Sample 1 CBDVA 0.13 4.1 CBGA 0.57 4.2 Sample 2 CBDVA <DL  CBGA 1.13 4.5 Sample 3 CBDVA <DL  CBGA 0.57 5.9 Sample 4 CBDVA 0.06 5.1 CBGA 1.14 3.3 Sample 5 CBDVA 0.09 1.6 CBGA 0.77 1.8 Sample 6 CBDVA <DL  CBGA 1.30 2.7 Sample 7 CBDVA 0.04 4.2 CBGA 1.40 1.3   <DL: data below the detection limit 2.3.2.4 Intermediate precision The quantitative data from the four replicates on the three days of analysis were used to determine the within-day, between-day and total standard deviations for calculating intermediate precision of the method. Intermediate precision ranged from 2.07 to 11.67 % RSD, summarized in Table 2.9. The HorRat ratios used to determine the acceptability of the % RSDs based on concentration ranged from 0.5 to 2.0, which is acceptable as specified by AOAC International guidelines (AOAC, 2013; Horwitz, 1982).  2.3.2.5 Recovery Cannabinoid recovery was evaluated for the four major cannabinoids: CBDA, THCA, CBD and THC. Three concentration levels were evaluated to represent a high, medium and 45 low concentration material with stinging nettle as the matrix blank. Recoveries are summarized in Table 2.11 and are within the acceptable ranges as specified by AOAC guidelines (AOAC, 2013). Table 2.11 Accuracy for the quantitation of major cannabinoids using spike recovery and stinging nettle as the matrix blank.  CBDA CBD THCA THC Sample ID Conc. (%w/w) Recovery (%) Conc. (%w/w) Recovery (%) Conc. (%w/w) Recovery (%) Conc. (%w/w) Recovery (%) High 3.5 97.7 1.0 91.3 3.5 96.1 1.0 90.7 Medium 0.4 92.6 0.4 95.5 0.4 90.7 0.4 96.2 Low 0.1 95.4 0.1 95.3 0.1 97.3 0.1 99.2  2.3.2.6 Limits of detection and quantitation The method detection limit (MDL) and limit of quantitation (LOQ) were determined using EPA’s method detection limit procedure (EPA, 2002). A test sample extract was diluted to very low concentrations to account for issues with closely eluting compounds. The detection and quantitation limits for each cannabinoid are summarized in Table 2.12. The limits for CBD are much higher in comparison with the other cannabinoids because of the number of close eluting peaks in the chromatogram. Most other cannabinoids have sufficient resolution from other peaks which do not impact their quantitation and detection. Table 2.12 Method detection limit and limit of quantitation of cannabinoids in solution and their respective concentrations in dried flowers using the EPA MDL procedures. Cannabinoid MDL LOQ Conc. (ppm) Amt. in sample (%w/w) Conc. (ppm) Amt. in sample (%w/w) CBDA 0.17 0.01 0.47 0.03 THCV 0.19 0.01 0.53 0.03 CBD 1.01 0.06 2.74 0.17 CBG 0.31 0.02 0.84 0.05 CBN 0.16 0.01 0.43 0.03 THCA 0.26 0.02 0.69 0.04 THC 0.09 0.01 0.25 0.02 CBC 0.34 0.02 0.93 0.06  46 2.4 Discussion With the rapid expansion of labs analyzing Cannabis, it is essential to have robust, versatile analytical methods. The currently available methods have several limitations. For example, resolution of the minor cannabinoids in some cases has been achieved only by selecting a less sensitive UV wavelength to achieve baseline resolution, while this reduction in sensitivity would impact the quantitation of low level cannabinoids found in many materials (Swift et al., 2013; Upton et al., 2014). Some methods fail to provide sufficient method development information to explain extraction solvent selection, extraction times, potential losses, degradation or inefficiencies (Swift et al., 2013; De Backer et al., 2009; Mehmedic et al., 2010). Many methods use chlorinated solvents with potential negative health and environmental impacts. Other methods use high pH mobile phases with pH > 5.0, which is above the pKa of the cannabinolic acids (De Backer et al., 2009). This elution system caused significant peak tailing and asymmetry when assessed during this project. The demand for cost-efficient quantitative methods for cannabinoids is growing. Many laboratories engaged in this work lack the expertise to work with advanced analytical instrumentation, such as mass spectrometric detectors. In this case, these detectors would impose a significant cost increase in the infrastructure and expertise. Mass spectrometry (MS) would allow for improved detection limits, selectivity and sensitivity, but given the performance of this method with UV absorbance and the high concentrations of the major cannabinoids, the use of MS detection is not necessary to increase method fitness. Care was taken to ensure that the mobile phases used are MS compatible for those with the instrument capabilities who require improved sensitivity and putative identification of additional minor cannabinoids. This method is an improvement over previous methods that can be used in a variety of settings and has the potential to be expanded for inclusion of new cannabinoids as required. To date, many jurisdictions only require the quantity of total THC and total CBD in the 47 products, while with improvements in analytical instrumentation, columns and detection techniques, the ability to expand regulations to acids, neutral forms, and minor cannabinoids is straightforward. There is considerable concern around the use of GC for quantitation of cannabinoids using in-injector decarboxylation because of conversion issues, which can reduce the accuracy and precision (UNODC, 2009). This issue is no longer a concern when quantifying cannabinolic acids separately from neutral cannabinoids. The information of acid content is also important for those that do not smoke Cannabis as the pharmacology of acids varies compared to neutral cannabinoids (Burstein, 1999). Understanding the cannabinoid profiles of different Cannabis strains will allow additional information for clinical researchers to understand the complex composition of these plants and their roles in pain regulation and treatment of a variety of other illnesses. This optimized HPLC-DAD method has a reduced extraction time, uses greener solvents, which can be adopted by other Cannabis testing laboratories. Sample turnaround is reduced, while method validation confirmed that the method produced repeatable, accurate results. The sample preparation eliminates the use of chloroform, which has been routinely used in cannabinoid analysis, reducing material costs, use of greener solvents and improved laboratory safety for personnel. This method can be used in a variety of settings from clinical studies, research, quality control and regulatory evaluation of this growing industry.   48 Chapter 3: Chemometric Analysis of Cannabinoids: Chemotaxonomy and Domestication Syndrome 3.1 Synopsis Cannabinoids are produced by the condensation of geranyl pyrophosphate (GPP) and olivetolic acid, products of the methylerythritol phosphate (MEP) and polyketide pathways, respectively (Flores-Sanchez & Verpoorte, 2008). This condensation reaction occurs by the prenylase geranyl diphosphate: olivetolate geranyltransferase (GOT) enzyme to produce cannabigerolic acid (CBGA). Synthesis occurs in the glandular trichomes which concentrate on female inflorescences. CBGA is the intermediate used by THCA synthase, CBDA synthase and CBCA synthase to produce THCA, CBDA and CBCA, respectively (Taura et al., 1996; Taura et al., 1995; Morimoto et al., 1999). These cannabinoids are the predominant form in plants and are decarboxylated to THC, CBD and CBC during processing, heating and storage. GPP can condense with polyketides that have different sidechain lengths, such as divarinolic acid, to produce other cannabinoids such as THCVA and CBDVA.  The domestication of Cannabis has included human selection, inbreeding and cross breeding as well as natural outcrossing and genome mixing (Clarke & Merlin, 2016b). Large genetic variance has been observed between C. indica, C. Sativa and identically named strains (Sawler et al., 2015; Soler et al., 2017). Controlled breeding programs to develop varieties or cultivars with unique phytochemical profiles have been limited (Small, 2015; De Meijer, 2014). With several hundred or perhaps thousands of strains of Cannabis currently being produced in legal and illegal markets, there is a possibility of identical or very closely related strains being sold with different names by different producers. Providing total THC and CBD may not provide sufficient information to distinguish strains, therefore metabolomic approaches could be used to predict the cannabinoid composition to understand the complex 49 phytochemistry of Cannabis. To investigate these hypotheses, Cannabis strains sold by licensed producers primarily based on total THC/CBD content were collected and analyzed for known cannabinoids using the validated method in Chapter 2 to establish clusters of similar plant materials. An untargeted metabolomics approach was used to identify previously uncharacterized compounds and chemical relationships. Five clusters of chemotaxonomically indistinguishable strains were identified within the collection.   3.2 Experimental 3.2.1 Reagents Methanol, acetonitrile, ammonium formate and formic acid (98%) were HPLC grade. Water was deionized and purified to 18.2 MΩ using a Barnstead Smart2Pure nanopure system (Thermo Scientific). Cannabinoid standards for quantification were purchased from Cerilliant Corp. (Round Rock, TX) for THCA, Δ9-THC, CBDA, CBD, CBG, CBC, THCV and CBN, Δ8-THC, CBDVA, CBDV, CBGA and CBL. All standards were provided as 1.0 mg/mL solutions in either methanol or acetonitrile. 3.2.2 Test materials Thirty-three strains of Cannabis were purchased from five licensed producers in Canada under the Access to Cannabis for Medical Purposes Regulations, and laboratory analysis was performed under a Health Canada Controlled Drugs and Substances License. The test samples were provided as whole or milled flowers in 5, 10 and 15 gram packages and stored at room temperature until use. Due to the legal restrictions pertaining to the storage of Cannabis strains, submission of voucher specimens to a herbarium were not possible, but given the regulatory framework controlling commerce and access and direct purchasing from licensed producers, the identity of the strains has been unambiguously confirmed as Cannabis sativa L. 50 3.2.3 Targeted metabolomics of cannabinoids The content of 13 cannabinoids was determined according to the validated analytical method as described in Chapter 2. In brief, ground Cannabis flowers (0.200 g) were extracted with 25.0 mL of 80% methanol in a 50 mL amber centrifuge tube for 15 minutes by sonication at room temperature and vortexed every 5 minutes, followed by centrifugation at 4500 g for 5 minutes and filtered with a 0.22 µm PTFE filter. Extracts were diluted to within the calibration range using the extraction solvent and placed in the 4 °C sample holder for same-day analysis. Chromatographic separation was performed on an Agilent 1200 RRLC with a Kinetex C18 100 mm x 3.0 mm, 1.8 µm column (Phenomenex; Torrance, CA) using gradient elution with 10 mM ammonium formate (pH 3.6) and acetonitrile. The autosampler was maintained at 4 °C and detection was at 220 nm. The peak areas for peaks with acidic or neutral cannabinoid UV spectra eluting between 2.5 and 14.5 minutes were collected using Chemstation software (Agilent Technologies) and known cannabinoids were identified. Known cannabinoids were quantified in % w/w against their individual calibration curves using external calibration in Microsoft Excel™. The total THC content was determined as the sum of THC and THCA in addition to the total degradation products of THC, CBN and Δ8-THC, adjusted by the molar mass ratios. CBD content was determined as the sum of CBDA and CBD adjusted by molar mass ratios. 3.2.4 Untargeted metabolomics Unknown cannabinoids were identified and numbered in sequential order as they appeared in the chromatogram. Unknown cannabinoids were quantified as THC equivalents using the THC calibration curves and ordered in sequential order in the chromatogram as CMPD#. 51 3.2.5 Data analysis For multivariate analysis, missing values were replaced with the method detection limit (MDL) divided by 2 for each assigned cannabinoid. In the cannabinoid profiles, where the MDL has not been determined for unassigned peaks, the missing data was replaced with half of the MDL of THC. Pearson correlation coefficients to determine relationships between metabolites were calculated using the cor script in R. As the concentration of a given metabolite does not necessarily correlate with pharmacological activity, the data were autoscaled by mean centering and scaling to unit variance in order to give each metabolite equal weight prior to multivariate analyses. Principal component analysis (PCA) and multiple linear regression (MLR) analysis were subsequently modeled using Solo+MIA (Eigenvector Research). 3.3 Results 3.3.1 Targeted metabolomics of cannabinoids Two cannabinoids for which standards were obtained, CDBV and CBL, were not detected in any strain. The 11 remaining cannabinoids with available chemical reference standards were identified and quantified. THCA content ranged from 0.76 to 20.71% w/w, with almost a linear increase in content from the lowest to highest strain with an r2 of 0.97, while CBDA content ranged from <MDL to 18.11 % w/w, with the highest CBDA strains having the lowest THCA contents (Figure 3.1). In THC abundant strains the CBDA levels were less than 0.15%, while in CBD abundant strains the content was greater than 5%. THC, the decarboxylated form of THCA, was present in strains from <LOQ up to 2% by weight in some strains, while CBD contents ranged from <MDL to 0.8%. CBD was most prevalent in high CBDA strains. In addition, 7 cannabinoids present at lower levels were quantified using individual calibration standards: THCV, CBG, CBN, CBC, CDBVA, CBGA and Δ8-THC. 52  Figure 3.1 Biosynthetic pathway of cannabinoids originating from olivetolic acid and geranyl pyrophosphate. Graphs describe the cannabinoid contents within the 33 strains obtained arranged from lowest to highest total THC. Error bars represent SEM, n=3. Order of strains from left to right are: can34, 38, 30, 35, 19, 16, 14, 39, 41, 31, 28, 40, 23, 33, 12, 22, 10, 20, 37, 42, 11, 25, 13, 27, 36, 24, 21, 29, 15, 18, 32, 17, 26.Olivetolic Acid Geranyl pyrophosphate (GPP)CBCACBGACBGTHCA CBDACBDTHCCBCTHCA Synthase CBDA SynthaseDecarboxylationGOT02468101214161820can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.511.522.53can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.10.20.30.40.50.60.70.80.9can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.050.10.150.20.250.3can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)0510152025can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.511.522.5can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)DecarboxylationDecarboxylation53 3.3.2 Classification of strains It was hypothesized that individual plant breeders selected for Cannabis strains by up-regulating and down-regulating specific enzymes within the biosynthetic pathways resulting in a redirection of metabolites between THC and CBD. The data analysis identified 5 clusters of strains that fall within a narrow range of CBD/THC values consistent with this hypothesis (Table 3.1). The branch of the biosynthetic pathway with olivetolic acid and geranyl pyrophosphate as precursors produces CBGA, CBG, CBCA, CBC, THCA, THC, CBDA and CBD (Figure 3.1). Strains from all clusters had measurable amounts of CBGA, CBG, THCA and THC (Figure 3.1). Nine strains from the clusters with higher concentrations of THCA (blue and purple) did not contain detectable levels of CBC (Figure 3.1). Two of the clusters did not contain significant quantities of CBDA and CBD (Figure 3.1; blue and purple). One strain was different from all others with a greater abundance of CBDA and detectable levels of CBGA, CBG CBC, and CBD with minimal amounts of THCA and THC (Figure 3.1; red).  Table 3.1 Strains of Cannabis were clustered into 5 distinct groups that could be separated by the flow of metabolites through the CBD and THC pathways. Group Colour Code CBD Range (% w/w) THC Range (% w/w) # Strains A Blue <MDL – 0.08 11.3 – 19.1 20 B Purple <MDL – 0.02 8.0 – 9.9 3 C Orange 7.1 – 9.7 5.0 – 6.7 6 D Green 5.3 – 8.8 1.7 – 3.1 3 E Red 16.1 0.7 1  Compounds produced from the precursors divarinolic acid, a polyketide with a propyl side chain, and geranyl pyrophosphate via CBGVA were also found to differ by strain cluster (Figure 3.2). CBGVA appears to be a branch point for the allocation of resources in Cannabis between THCV and CBDVA indicating that the enzyme activity or the resource allocation mechanism for production of CBDVA was lost in the breeding process of strains clustered in the purple and blue groups (Figure 3.2). 54  Figure 3.2 Biosynthetic pathway of cannabinoids originating from divarinolic acid and geranyl pyrophosphate. Graphs describe the cannabinoid contents within the 33 strains obtained arranged from lowest to highest total THC. Error bars represent SEM, n=3. X’s indicate a biosynthetic break in the pathway.  Divarinolic acid Geranyl pyrophosphate (GPP)CBGVATHCVA CBDVATHCVCBDVCBCVACBGV CBCV00.020.040.060.080.10.120.140.16can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)00.10.20.30.40.50.60.70.80.91can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)55 3.3.3 Untargeted metabolomic analysis In addition to the 11 cannabinoids that corresponded with authentic standards, 21 peaks were identified in the chromatograms with UV spectra characteristic of cannabinoids. By comparison to THC, the contents were estimated from <MDL up to 0.34 % by weight. Two unknown cannabinoids (CMPD7 and CMPD11) were detected in all strains, while CMPD3 and CMPD20 were each only detected in a single strain. 3.3.4  Relationships between known and unknown cannabinoids Principal component analysis (PCA) of the autoscaled cannabinoid data was plotted to show the clustering of the samples in an unsupervised fashion (Figure 3.3). In the PCA plot, the first two principal components (PC) captured 36.6% of the variance in the data. Based on the loadings plot (Figure 3.4), the first PC was strongly influenced by the THCA and CBDA content of the strains, which are negatively correlated. There are two high THC strains (CAN17 and CAN21) and one CBD strain (CAN34) that were outside the 95% confidence limit of the total data variance. Based on the loadings plot (Figure 3.4), the high THC strains may be influenced by a significant number of low abundance cannabinoids including CBGA, CMPD12, and CMPD11. The CBD strain (red) outside of the 95% confidence limit is likely due to its significantly higher CBDA content and less than 1% total THC.  56  Figure 3.3 PCA scores plot of cannabinoid profiles classified according to THC/CBD contents. Strains prepared in triplicate, analyzed as individual data points and color coded according to cannabinoid classification.   Figure 3.4 PCA loadings plot of cannabinoid profiles classified according to THC/CBD contents.  57   While the first two PCs of PCA describe 36% of the variance, there is a remaining 64% of the variance in the cannabinoids not being described by the model. Therefore, additional models were employed to understand the relationships between cannabinoids and to identify additional strain classes based on the content of these 32 different cannabinoids. Multiple linear regression (MLR) analysis showed that 14 cannabinoids were better suited compared to all cannabinoids for predicting THCA content with validation r2 values improving from 0.18 (Figure 3.5) and 0.88 (Figure 3.6), respectively and for predicting CBDA content 14 cannabinoids improved the validation r2 values from 0.006 (Figure 3.7) to 0.92 (Figure 3.8) when compared with using the entire data set.  Figure 3.5 Multiple linear regression model to estimate the THCA content from the entire cannabinoid dataset. (n=3) 0 5 10 15 20 25-40-30-20-1001020304050Y Measured 1 THCAY CV Predicted 1 THCARMSEC = 0.45372RMSECV = 14.5945Calibration Bias = 1.7764e-15CV Bias = -2.6047R 2^ (Cal,CV) = 0.994, 0.18358  Figure 3.6 Improved MLR model to estimate the THCA content from a reduced data set of 14 cannabinoids. (n=3)   Figure 3.7 MLR model to estimate CBDA content from the entire cannabinoid dataset. (n=3) 0 5 10 15 20 25-50510152025Y Measured 1 THCAY CV Predicted 1 THCARMSEC = 1.146RMSECV = 1.9615Calibration Bias = -3.5527e-15CV Bias = 0.11194R 2^ (Cal,CV) = 0.960, 0.8870 2 4 6 8 10 12 14 16 18 20-40-30-20-100102030Y Measured 1 CBDAY CV Predicted 1 CBDARMSEC = 0.1453RMSECV = 9.4984Calibration Bias = -3.5527e-15CV Bias = -0.78452R 2^ (Cal,CV) = 0.999, 0.00659  Figure 3.8 Improved MLR model to estimate the CBDA content from a reduced dataset of 14 cannabinoids. (n=3)     Pearson correlations were used to determine whether any of the unidentified cannabinoids could be associated with the major cannabinoids THCA, THC, CBDA and CBD (Table 3.2). There was no significant correlation of THCA or THC and any of the unknown compounds (Table 3.2). The CBDA content was positively correlated with CMPD1, CBDVA, CMPD5, CMPD6, CMPD16 and CMPD18 (Table 3.2) with ρ ranging from 0.61 to 0.93. CBD was potentially weakly correlated with CMPD1, CMPD6 and CBDA (Table 3.2). The correlations between all detected cannabinoids are described in Figure 3.9.   0 2 4 6 8 10 12 14 16 18 20-50510152025Y Measured 1 CBDAY CV Predicted 1 CBDARMSEC = 0.88148RMSECV = 1.3682Calibration Bias = 1.3323e-15CV Bias = 0.099394R 2^ (Cal,CV) = 0.964, 0.91860 Table 3.2 Pearson correlation coefficients of all cannabinoids relative to the four major cannabinoids (THCA, CBDA, THC and CBD). Cannabinoid THCA CBDA THC CBD CMPD1 -0.71 0.93 -0.39 0.54 CMPD2 0.22 -0.14 0.03 -0.11 CBDVA -0.70 0.93 -0.36 0.49 CMPD3 -0.20 0.20 0.15 0.40 CMPD4 0.18 -0.19 0.17 -0.20 CMPD5 -0.36 0.61 -0.26 0.12 CMPD6 -0.65 0.84 -0.27 0.58 CBDA -0.81 1.00 -0.34 0.68 CMPD7 0.53 -0.41 0.21 -0.29 CMPD8 0.21 -0.26 0.12 0.01 CBGA 0.46 -0.18 0.52 -0.17 CMPD9 -0.29 0.22 0.23 0.30 CMPD10 0.16 -0.12 0.35 -0.12 THCV 0.05 0.16 0.15 -0.10 CMPD11 0.36 -0.09 0.31 -0.21 CBD -0.68 0.68 0.00 1.00 CMPD12 0.28 -0.23 0.24 -0.12 CBG 0.66 -0.35 0.43 -0.27 CMPD13 -0.04 0.02 0.23 0.17 CMPD14 -0.19 0.34 0.07 0.20 CMPD15 -0.24 0.39 -0.32 0.08 CMPD16 -0.55 0.68 -0.08 0.44 CMPD17 0.00 0.19 -0.19 0.07 THCA 1.00 -0.81 0.39 -0.68 CBN -0.26 0.32 -0.05 0.41 CMPD18 -0.73 0.91 -0.25 0.72 THC 0.39 -0.34 1.00 0.00 8-THC 0.28 -0.14 -0.13 -0.07 CBC -0.21 0.30 0.63 0.59 CMPD19 0.19 -0.08 -0.24 -0.21 CMPD20 -0.04 -0.11 0.01 -0.10 CMPD21 0.02 0.05 -0.05 0.18  61  Figure 3.9 Heatmap illustrating the Pearson correlation coefficients for all of the cannabinoids detected using HPLC-UV separation.   3.3.5 Putative identifications and pathways Ten of the unidentified cannabinoids were found across multiple strains from all clusters (Figure 3.10). CMPD1 was strongly correlated with CBDA according to Pearson’s correlation (Table 3.2) and although it was found in many of the strains classified as blue or 62 purple, it was at much higher levels in the red, green and orange clusters (Figure 3.11a). Compounds 3,5,6,15 and 18 were found only in the CBD-rich clusters red, green and orange (Figure 3.11b,c,d,e,and f). Compounds 2, 12, and 20 were found only in THC-dominant strains (Figure 3.12). 00.010.020.030.040.050.060.070.08can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(a)00.050.10.150.20.25can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(b)00.020.040.060.080.10.120.14can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(c)63   00.010.020.030.040.050.060.07can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(d)00.050.10.150.20.25can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(e)00.020.040.060.080.10.120.140.160.18can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(f)00.020.040.060.080.10.120.14can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(g)64    Figure 3.10 Cannabinoid content of the unidentified cannabinoids found throughout the sample set. (a) CMPD4, (b) CMPD7, (c) CMPD8, (d) CMPD9, (e) CMPD10, (f) CMPD11, (g) CMPD14, (h) CMPD16, (i) CMPD19, (j) CMPD21. Error bars represent SEM, n=3.  00.050.10.150.20.250.30.350.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(h)00.010.020.030.040.050.060.070.08can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(i)00.0050.010.0150.020.0250.030.035can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(j)65    00.050.10.150.20.250.3can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(a)00.0050.010.0150.020.0250.030.0350.04can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(b)00.0020.0040.0060.0080.010.012can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(c)00.010.020.030.040.050.06can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(d)66   Figure 3.11 Cannabinoid content of the unidentified cannabinoids found in CBD-rich strains. (a) CMPD1, (b) CMPD3, (c) CMPD5, (d) CMPD6, (e) CMPD15, (f) CMPD18. Error bars represent SEM, n=3.    00.010.020.030.040.050.06can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(e)00.020.040.060.080.10.120.14can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(f)67   Figure 3.12 Unidentified cannabinoids found only in THC-dominant strains. (a) CMPD2, (b) CMPD12, (c) CMPD20. Error bars represent SEM, n=3.  3.4 Discussion The long history of use by humans has made the exact region of origin for Cannabis difficult to establish, though literature supports Northeast Asia (Clarke & Merlin, 2016b; Small, 2015). Breeding of Cannabis cultivars in the hemp industry has focused on morphological improvements through established breeding programs, while marijuana, or drug-type Cannabis, has primarily taken place in underground/clandestine programs through crossing 00.050.10.150.20.250.3can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(a)00.020.040.060.080.10.12can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(b)00.0050.010.0150.020.0250.030.035can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Concentration (%w/w)(c)68 landraces and/or “indica” and “sativa” lineages with the major focus on increasing the yield of THC (Clarke & Merlin, 2016b; Small, 2015). The genetic diversity between marijuana strains is lower in comparison with hemp varieties due to crossing closely related varieties (Sawler et al., 2015; Soler et al., 2017). CBDA and THCA synthases are thought to be controlled by two alleles on a single locus (B), crossing of CBD and THC-dominant strains will produce offspring with intermediate THC:CBD ratios (De Meijer et al., 2003; De Meijer 2014). With the prevalence of propagation through cuttings of mother plants, feminization of seeds and production of sensimilla, the need for male plants has decreased resulting in potential loss of genetic and phytochemical diversity (Clarke & Merlin, 2016b). The ‘sativa’ and ‘indica’ lineages used to describe Cannabis throughout the industry are based on postulation that sativa strains originated from European hemp cultivars, while indica are from potent, resinous Indian Cannabis (Small, 2015) but given the use and trade of the plant in ancient times, the exact origin is unknown, and these may not be distinct species. Modern strains are considered dominant in either of these two “lineages” or hybrids between close relatives. These classifications focus on the pharmacological effects associated with the strains where sativa plants are considered stimulating and indica plants are associated with relaxation and sedation, but this is not a botanical or chemotaxonomical classification. Comparisons of cannabinoid contents of these classifications have shown that the THC content can be identical between these two classification groups (Soler et al., 2017; Hazekamp & Fischedick, 2012).  Many questions remain open. What is a “strain”? Does a “strain” represent a phytochemically unique variety? Are “strains” from different growers different? Is there a more appropriate way to classify “strains”? Are these cultivars, varietals, landraces or even species? What is the impact of domestication on the ecological fitness of the species? Breeding closely related plants potentially leads to loss of genetic diversity within the genome (Meyer et al., 2012). Traits signifying domestication syndrome include phenotypic 69 changes such as increased seed size, loss of shattering, changes in reproduction, changes in secondary metabolites and loss of pest resistance compared with wild ancestors (Meyer et al., 2012; McKey et al., 2010). Reviews of Cannabis breeding have summarized domestication in terms of morphology, while focus on secondary metabolism has primarily been on THC content (Clarke & Merlin, 2016b; Small 2013). Recent forensic evaluations of confiscated sensimilla Cannabis in the US has shown dramatic increases in THC content over the last 30 years, from 6.3% to 11.5% (ElSohly et al., 2000; Mehmedic et al., 2010), while strains with greater than 20% THC are available in the marketplace. This artificial increase in THC production has resulted in the loss of CBDA synthase activity in THC-dominant strains. Although crossbreeding will result in THC:CBD hybrid offspring, the loss of other biosynthetic pathways is unknown due to the non-rigorous breeding programs focusing primarily on the production of a single metabolite. The cannabinoid profiles observed in this study indicate that these breeding programs have also impacted unknown related metabolites with undetermined function.  Metabolomic analysis can generate chemotaxonomic classifications of plants in addition to hypothesis-generating insight of data correlations, metabolite identification and relationships that would not be possible through single metabolite evaluation (Turi et al., 2015; Fiehn, 2002). Using the correlation data, PCA loadings plots and UV spectra, the putative identity of some of these unknown cannabinoids were determined. For example, CMPD6 had a Pearson correlation coefficient of 0.89 with CBDA and occupies the same space within the PCA loadings plot. The UV spectra with a maximum of 224 nm identifies this compound as an acidic cannabinoid which was only detected in the presence of CBDA (Figure 3.13). Further evaluation showed that it eluted between CBDVA and CBDA, therefore is hypothesized to be CBDA-C4 with a butyl side chain on the polyketide (Table 3.3) (Smith, 1997). Likewise, CBDA-C1, CBDM, and CBDMA were putatively identified based on their UV spectra and 70 chromatographic elution order which provides information on hydrophobicity, and correlation with CBDA (Figure 3.13, Table 3.3). Due to the presence of THCA synthase in all strains, the correlation of THC-type cannabinoids is less obvious. It was previously reported that low abundance cannabinoids may be regulated by upstream biosynthesis of precursor polyketides (Shoyama et al., 1984). There were fewer unknown cannabinoids in the strains selected for higher THC content. With such strong emphasis on the synthesis of a single metabolite there is a strong possibility that other biosynthetic pathways have been lost in the process (Meyer et al., 2012; McKey et al., 2010).   Figure 3.13 UV spectra of acidic and neutral unidentified cannabinoids in comparison to known cannabinoids. (a) CMDP1, (b) CBDA, (c) CMPD18, (d) CBD.    Wavelength (nm)200 220 240 260 280 300 320 340 360 380Absorbance (mAU)0246810Wavelength (nm)200 220 240 260 280 300 320 340 360 380Absorbance (mAU)0100200300400500600Wavelength (nm)200 220 240 260 280 300 320 340 360 380Absorbance (mAU)01234567Wavelength (nm)200 220 240 260 280 300 320 340 360 380Absorbance (mAU)02.557.51012.51517.5(a) (b)(c) (d)71 Table 3.3 UV spectrum of known and unknown cannabinoids along with their putative identification based on UV spectrum, elution order and correlations.  Cannabinoid Retention Time (min) UV Spectrum: Acidic/Neutral Putative ID CMPD1 2.6 Acidic CBDA-C1 CMPD2 2.9 Acidic  CBDVA 4.3 Acidic  CMPD3 4.4 Acidic  CMPD4 4.8 Acidic  CMPD5* 5.4   CMPD6 5.6 Acidic CBDA-C4 CBDA 6.9 Acidic  CMPD7 7.4 Acidic  CMPD8 7.7 Acidic  CBGA 7.9 Acidic  CMPD9* 8.2   CMPD10 8.5 Neutral  THCV 8.7 Neutral  CMPD11 8.9 Neutral  CBD 9.1 Neutral  CMPD12 9.3 Acidic  CBG 9.4 Neutral  CMPD13 10.2 Neutral  CMPD14 10.5 Neutral  CMPD15 10.7 Acidic CBDMA CMPD16 10.8 Neutral  CMPD17* 10.9   THCA 11.2 Acidic  CBN 11.3 Neutral  CMPD18 11.7 Neutral CBDM THC 12.3 Neutral  8-THC 12.6 Neutral  CBC 13.6 Neutral  CMPD19 13.9 Neutral  CMPD20 14.0 Acidic  CMPD21 14.4 Neutral  *Acidic/Neutral designations of CMPD5, CMPD9 and CMPD17 are unknown Cannabinoids have varying pharmacological effects. THC is known primarily for its psychoactivity but is also a suitable anti-nausea drug for chemotherapy patients (Plasse et al., 1991; Pertwee, 2008). Unfortunately, it has several side effects that are reduced in the presence of CBD (McPartland et al., 2015). THC has also been shown to have anticonvulsant, anti-inflammatory activities and improvements in multiple sclerosis, pain, anorexia, diabetes and metabolic syndrome (Russo, 2011). On the other hand, CBD has been shown to enhance 72 the efficacy of THC while suppressing psychosis, anxiety and depression (Russo et al., 2005). CBD also has anticonvulsant and anti-inflammatory activities (Consroe et al., 1961; Klein, 2005). Although these are the two major cannabinoids, several minor cannabinoids have been investigated for therapeutic potential. CBG was shown to have strong potential as a therapeutic in inflammatory bowel disease due to its strong anti-inflammatory activity, while it was also shown to counteract the anti-nausea effects of CBD (Borelli et al, 2013; Rock et al., 2011). THCV is anticonvulsant, anti-inflammatory, can reduce inflammatory pain and reduces nausea (Rock et al., 2013; Bolognini et al., 2010; Hill et al., 2010). CBDV has anticonvulsant and antinausea activities (Rock et al., 2013; Amada et al., 2013). CBC has anti-inflammatory and antidepressant activities (Izzo et al., 2010; El-Alfy et al., 2010). While these are the cannabinoids that have been identified in this study, there are an additional 21 cannabinoids which were not identified. This indicates the significance of understanding the composition of these metabolites within single strains, which can eventually be evaluated through pre-clinical and clinical studies coupled with chemometric analysis to evaluate the interactions between these metabolites and their pharmacological significance. Several classification systems have been proposed for Cannabis based on a limited number of phenotypic attributes (Clarke & Merlin, 2016b; Small, 2015; Small & Beckstead, 1973). The concept of a “strain” does not reflect the crop domestication, breeding programs or plant chemistry. The strains available in the Canadian marketplace are closely related and evaluating single metabolite classes does not provide sufficient information to understand the phytochemical diversity available. The abundance of secondary metabolites within plants does not necessarily correlate to pharmacological significance and with Cannabis there is the postulated “entourage effect” describing the synergistic effects of many metabolites for anecdotal medicinal efficacy (Russo, 2011).  Domestication of the crop has limited the genetic variability in the crop and the impact on crop diversity, physiology and metabolism is not fully 73 understood. Further research is needed to evaluate the low abundance cannabinoids for medicinal efficacy and to determine their roles in plant metabolism.     74 Chapter 4: Domestication Syndrome and Metabolomics of Volatile Constituents in Cannabis 4.1 Synopsis Besides cannabinoid potency, clandestine breeding programs also rely heavily on other visual and organoleptic cues when estimating the quality and marketing potential of new strains. This included several aspects such as aroma, plant morphology, color, bud size, yield, vigor, etc. (Clarke & Merlin, 2016b; Small, 2015). The aroma of strains plays a significant role in the selection, preference and quality indications, which may result in strains which are not genetically different from one another (Gilbert & DiVerdi, 2018). Genetic variation within strains is highlighted by the expression of the phytochemicals present and as monoterpenes and sesquiterpenes are volatile constituents, their aromas may have contributed to selections in breeding.  Monoterpenes and sesquiterpenes are produced by the methylerythritol phosphate (MEP) and mevalonic acid (MVA) pathways, respectively. Both pathways independently produce isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) in different cell regions (Tholl, 2006). Geranyl pyrophosphate (GPP) is the precursor for C10 monoterpenes. GPP is produced by the condensation of IPP and DMAPP with geranyl pyrophosphate synthase enzymes. Farnesyl pyrophosphate (FPP) requires 2 IPP and 1 DMAPP via farnesyl pyrophosphate synthases, which is the precursor for C15 sesquiterpenes (Tholl, 2006). Different terpene synthases then use these precursors to produce the monoterpenes and sesquiterpenes found in plants (Booth et al., 2017). In Cannabis, these biosynthetic reactions take place in the glandular trichomes, where the terpenes are stored. When trichomes are broken, the terpenes volatilize to produce the aromas of Cannabis strains. 75 It was hypothesized that Cannabis breeders selected for scent notes that were believed to be associated with specific phytochemical profiles and in this process,  they selected for modified terpene biosynthesis. To investigate this hypothesis, the collection of 33 Cannabis strains evaluated for cannabinoid variation, described in Chapter 3, were profiled for monoterpenes and sesquiterpenes using headspace GC-MS. A combination of metabolomic evaluations were performed to understand the impacts of breeding and aroma on Cannabis domestication. 4.2 Experimental 4.2.1 Test materials The thirty-three Cannabis strains purchased from five licensed producers that were used in Chapter 3 of this work were also used to evaluate the volatile constituents. The test samples were provided as whole or milled flowers in 5, 10 and 15 gram packages and stored at room temperature until use. 4.2.2 Reference standards Cannabis Terpene Mix A and Mix B containing 20 and 15 terpenes, respectively, at 2000 µg/mL in methanol were purchased from Sigma Aldrich (Oakville, ON, Canada). Cannabis Terpene Mix A contained: α-pinene, camphene, β-pinene, 3-carene, α-terpinene, limonene, γ-terpinene, fenchone, fenchol, camphor, isoborneol, menthol, citronellol, pulegone, geranyl acetate, α-cedrene, α-humulene, nerolidol, cedrol and α-bisabolol. Cannabis Terpene Mix B contained: β-pinene, 3-carene, p-cymene, limonene, terpinolene, linalool, camphor, borneol, α-terpineol, geraniol, β-caryophyllene, cis-nerolidol, β-eudesmol and phytol. All standards were stored according to the manufacturer’s recommendation until use. 76 4.2.3 Evaluation of volatile constituents Monoterpenes and sesquiterpenes were evaluated using headspace GC-MS, adopted from Giese et al. (2015). Cannabis flowers were ground with liquid nitrogen and immediately 100.0 mg was added to a 20 mL gas tight GC headspace vial and closed with a screw-top cap (Agilent Technologies, Mississauga, ON). Using a CTC Analytics Combi-PAL headspace autosampler, each vial was transferred to a heated incubator at 80 °C for 15 minutes and agitated at 500 rpm prior to analysis. The vial headspace was sampled (1000 µL) using a syringe at 120 °C and transferred to the GC inlet and injected. The injector temperature was 230 °C with at a split ratio of 10:1. GC analysis was undertaken on an Agilent 7890A GC coupled to a 5975B mass spectrometer (MS). Separation was achieved with a 20 m x 180 µm ID, 0.18 µm film thickness J&W DB-5MS column (Agilent Technologies). Helium was used as the carrier gas at a flow rate of 1.3 mL/min. The column was held at 50 °C for 3 minutes followed by a ramp to 170 °C at 5 °C/min for a total run time of 27 minutes. MS detection with electron impact ionization at 70 eV was used to collect mass spectra from m/z 50 to 500. The MS quadrupole and source temperatures were 230 °C and 150 °C, respectively. 4.2.4 Chemometrics 4.2.4.1 Metabolite profiling Terpenes were identified by comparison to the reference standard solutions and mass spectra using the National Institute of Standards and Technology (NIST) spectral database (NIST 11). Additionally, retention indices were compared to published literature to confirm elution order (Babushok et al., 2011). Multivariate curve resolution (MCR), using SOLO+MIA software, was employed to separate co-eluting monoterpenes and peak areas were determined using R (Ruckebusch & Blanchet, 2013). Peaks were manually aligned based on 77 compound identity and retention time using Microsoft Excel. Missing values were replaced with half of the lowest value in the data set. 4.2.4.2 Identification of metabolite relationships Individual terpenes were plotted according to their cannabinoid profiles to identify trends within the data sets and classify them into unique groups. Trends evaluated included: present primarily in THC-dominant strains, present primarily in CBD-THC hybrid strains, and other unique correlations. Correlations between terpenes were confirmed by evaluating Pearson correlation coefficients using the R program cor.  4.2.4.3 Multivariate classification The data were autoscaled by mean centering and scaling to unit variance in order to give each metabolite equal weight prior to multivariate analyses. Principal component analysis (PCA) was subsequently performed using Solo+MIA (Eigenvector Research).  4.3 Results 4.3.1 Terpene profiles A total of 67 terpenes were detected composed of 29 monoterpenes and 38 sesquiterpenes. Monoterpenes accounted for 87.1 to 99.5 % of the headspace profiles, with an average of 94.6 %. Sesquiterpenes accounted for the remaining 0.5 to 12.9%. Four strains had less than 1% sesquiterpenes while the average content was 5.4%. Terpenes were identified using a combination of methods to ensure accurate identities were assigned to each metabolite. A summary of the terpene retention times, identification methods and identities are summarized in Table 4.1. Identification included the use of two reference standard mixes that were purchased from Sigma Aldrich which contained a combined 36 terpenes. For those without reference standards, the mass spectra were 78 compared with the NIST mass spectral database and relative matches were used to compare the identities of all terpenes to their structure. Based on comparison to known terpenes and the spectral matches, matches above a relative match of 850 were sufficient to identify the compound. In some cases, there were specific strains with a higher prevalence of a specific terpene which improved identification. Retention indices were also used to compare expected retention order of the terpenes to aide in identification (Babushok et al., 2011). Those with low spectral matches were considered tentative identification and are italicized in Table 4.1.  Table 4.1 Identification of terpenes in Cannabis by reference standard mix (std. mix), mass spectral and retention index comparisons. Tentatively identified compounds are italicized. Retention Time (min) Relative Match Identification Method(s) Strain used Identification 3.719 818 NIST/retention index can23 santolina triene 3.85 960 NIST/retention index can17 α-thujene 4.012 960 std. mix can27 α-pinene 4.4 964 std. mix can24 camphene 5.04 951 NIST/retention index can32 sabinene 5.136 937 std. mix can27 β-pinene 5.583 910 NIST/retention index can25 β-myrcene 5.745 828 NIST/retention index can33 2-carene 5.97 881 NIST/retention index can17 α-phellandrene 6.1 934 std. mix can17 3-carene 6.286 940 std. mix can17 α-terpinene 6.518 908 std. mix  p-cymene 6.625 960 std. mix can36 limonene 6.7 939 NIST/retention index can17 β-phellandrene 6.946 848 NIST/retention index can36 trans-β-ocimene 7.226 975 NIST/retention index can42 cis-β-ccimene 7.493 951 std. mix can17 γ-terpinene 7.888 909 NIST/retention index can17 Z-sabinene hydrate 8.266 942 std. mix can17 terpinolene 8.34 904 std. mix can21 fenchone 8.495 887 NIST/retention index can21 p-cymenene 8.832 964 std. mix can36 linalool 9.263 956 std. mix can15 exo-fenchol 9.431 906 NIST/retention index can15 trans-2-pinanol 10.042 855 NIST database can21 α-fenchene  79 Retention Time (min) Relative Match Identification Method(s) Strain used Identification 10.254 870 NIST/retention index can36 camphene hydrate 10.783 907 std. mix can29 endo-borneol 11.117 895 NIST/retention index can17/64 terpinen-4-ol 11.584 890 std. mix can26 α-terpineol 15.533 791 NIST/retention index can26 α-cubenene 16.059 860 NIST/retention index can36 ylangene 16.237 814 NIST/retention index can26 copaene 16.671 744 NIST/retention index can27 β-cubebene 17.067 n/a NIST/retention index can18 β-elemene 17.306 967 std. mix can24 caryophyllene 17.37 920 NIST/retention index can36 α-santalene 17.647 835 NIST/retention index can26 γ-elemene 17.755 896 NIST/retention index can37 α-guaiene 17.915 783 NIST/retention index can36 α-gurjunene 18.085  unknown  sesquiterp-1 18.204 946 std. mix can24 humulene 18.288 780 NIST/retention index can39 alloaromadendrene 18.379 921 NIST/retention index can36 cis-β-farnesene 18.631 780 NIST/retention index can27 (Z,Z)-α-farnesene 18.759 864 NIST/retention index can26 γ-muurolene 18.853 800 NIST/retention index can26 α-amorphene 18.915 902 NIST/retention index can36 4,11-selinadiene 19.042 907 NIST/retention index can36 β-selinene 19.168 853 NIST/retention index can36 γ-gurjunene 19.221 911 NIST/retention index can26 α-selinene 19.427 959 NIST/retention index can27 α-bulnesene 19.685 696 NIST/retention index can35 germacrene A 19.752 883 NIST/retention index can26 valencene 19.815 800 NIST/retention index can23 δ-cadinene 19.889  unknown can26 α-gurjunene derivative 20.013 865 NIST/retention index can36 β-sesquiphellandrene 20.112 917 NIST/retention index can12 β-maaliene 20.185 897 NIST/retention index can26 guaia-3,9-diene 20.288 929 NIST/retention index can26 selina-3,7(11)-diene 20.457 828 NIST/retention index can29 cis-α-bisabolene 20.669 914 NIST/retention index can26 germacrene B 20.79 868 NIST/retention index can12 δ-Selinene 21.237 810 NIST/retention index can18 caryophyllene oxide 21.679 908 NIST/retention index can22 guaiol 22.173 932 NIST/retention index can22/56 8-epi-γ-eudesmol 22.962 835 std. mix can28 β-eudesmol 23.186 850 std. mix can28 bulnesol 80 4.3.2 Terpene profile by strain The classification system based on cannabinoid content described in Chapter 3 was used to identify relationships of different terpenes across these strain classifications. To assess the relationships between THC, CBD and the terpenes, each terpene was graphed according to THC content from lowest to highest and color coded to strain class. Twelve terpenes were ubiquitous across all strains (Figure 4.1). Three monoterpenes: limonene, β-myrcene and α-pinene were abundant in the majority of strains, while the two most abundant sesquiterpenes: β-caryophyllene and humulene ranged from 0.2 to 5.5% and 0.3 to 1.5% respectively. 00.511.522.533.54can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)camphene                      (a)0123456can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-caryophyllene (b)81 00.20.40.60.811.21.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)guaia-3,9-diene               (c)00.20.40.60.811.21.41.61.8can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-guaiene (d)00.20.40.60.811.21.41.61.8can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)humulene                    (e)0102030405060can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)D-limonene               (f)82    00.050.10.150.20.250.30.350.40.450.5can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-maaliene (g)010203040506070can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-myrcene                 (h)010203040506070can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-pinene                   (i)024681012141618can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-pinene                    (j)83  Figure 4.1 Terpene profiles identified as those present across the entire dataset. (a) camphene (b) β-caryophyllene (c) guaia-3,9-diene (d) α-guaiene (e) humulene (f) D-limonene (g) β-maaliene (h) β-myrcene (i) α-pinene (j) β-pinene (k) selina-3,7(11)-diene and (l) valencene. Error bars represent SEM, n=3.  Seventeen terpenes were found in strains all of the cannabinoid groupings, but not in all strains (Figure 4.2). β-cubebene was found in all strains except the very low THC, high CBD strain (Figure 4.2d). There were considerable correlations among the lower abundance sesquiterpenes with correlation coefficients above 0.8 as visually represented in Figure 4.3. Correlations were observed between γ-muurolene, copaene, β-cubebene, elemol, Germacrene A, guaia-3,9-diene, β-maaliene, γ-maaliene, seling-3,7(11)-diene, α- selinene and δ-selinene, for which many of these metabolites were observed in either all or almost all cannabinoid strains and clusters. 00.511.522.5can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)selina-3,7(11)-diene         (k)00.050.10.150.20.250.30.35can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)valencene              (l)84 00.050.10.150.20.250.30.35can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)endo-borneol      (a)00.010.020.030.040.050.060.070.08can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)camphene                    (b)00.0050.010.0150.020.0250.03can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)copaene                    (c)00.050.10.150.20.25can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-cubebene                 (d)85  00.511.522.533.54can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)exo-fenchol                 (e)00.10.20.30.40.50.6can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)fenchone                  (f)00.050.10.150.20.25can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)Germacrene A               (g)00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-gurjunene derivative      (h)86   00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)γ-gurjunene                 (i)00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)γ-muurolene                 (j)00.20.40.60.811.21.41.61.8can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)trans-2-pinanol              (k)00.050.10.150.20.250.30.350.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)z-sabinene hydrate              (l)87    00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)4,11-selinadiene (m)00.050.10.150.20.250.3can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-selinene                   (n)00.050.10.150.20.250.30.35can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-selinene (o)00.050.10.150.20.250.30.350.40.45can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peeak Area (%)α-terpineol                    (p)88  Figure 4.2 Terpene profiles identified as those present across the different cannabinoids classes, but not present in all strains. (a) endo-borneol (b) camphene hydrate (c) copaene (d) β-cubebene (e) exo-fenchol (f) fenchone (g) Germacrene A (h) α-gurjunene derivative (i) γ-gurjunene (j) γ-muurolene (k) trans-2-pinanol (l) z-sabinine hydrate (m) 4,11-selinadiene (n) α-selinene (o) β-selinene (p) α-terpineol (q) Ylangene. Error bars represent SEM, n=3. 00.020.040.060.080.10.12can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)ylangene                      (q)89  Figure 4.3 Pearson correlations between monoterpenes and sesquiterpenes within the Cannabis dataset.   Nine terpenes were present in THC-dominant strains (Figure 4.4) and four were found to be in strains identified as mid-range THC (Figure 4.5). Santolina triene (tentative identification) was one of two monoterpenes observed to have correlations with sesquiterpenes sesquiterp-1 (unidentified) and δ-cadinene all present in this grouping (Figure 4.3).  90   00.0020.0040.0060.0080.010.012can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-amorphene                    (a)00.020.040.060.080.10.120.14can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)2-carene (b)00.0020.0040.0060.0080.010.0120.0140.0160.0180.02can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)caryophyllene oxide        (c)00.0020.0040.0060.0080.010.0120.014can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-cubenene                  (d)91   00.0020.0040.0060.0080.010.0120.0140.016can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-elemene                    (e)00.0050.010.0150.020.0250.030.0350.040.045can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)γ-elemene                       (f)00.0050.010.0150.020.025can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)(Z,Z)-α-farnesene           (g)00.050.10.150.20.250.30.350.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)Germacrene B                 (h)92  Figure 4.4 Terpene profiles present primarily in THC-dominant strains. (a) α-amorphene (b) 2-carene (c) caryophyllene oxide (d) α-cubenene (e) β-elemene (f) γ-elemene (g) (Z,Z)-α-farnesene (h) germacrene B (i) β-sesquiphellandrene. Error bars represent SEM, n=3.    00.0010.0020.0030.0040.0050.0060.0070.008can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-sesquiphellandrene           (i)93  Figure 4.5 Terpene profiles for terpenes found predominantly in mid-level THC-dominant strains. (a) δ-cadiene (b) α-gurjunene (c) santolina triene (d) sesquiterp-1 (unidentified). Error bars represent SEM, n=3.    00.0050.010.0150.020.0250.03can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)δ-cadiene                    (a)00.0050.010.0150.020.0250.03can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-gurjunene               (b)00.10.20.30.40.50.6can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)santolina triene               (c)00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)unidentified sesquiterp-1     (d)94 Eighteen terpenes were present in high proportions in strains identified as very high THC and mid-level THC/CBD (Figure 4.6). In this grouping, terpinolene was the most dominant monoterpene, with less than 0.3% in 27 of the 33 strains, but ranges from 13.4 to 41.2% in the six strains with this distinctive monoterpene profile: CAN16, CAN17, CAN19, CAN21, CAN32 and CAN33. Terpinolene was correlated with other monoterpenes: α-thujene, α-phellandrene, 3-carene, α-terpinene, p-cymene, β-phellandrene, α-terpinene, and terpinen-4-ol with correlation coefficients ranging from 0.95 to 0.99 (Figure 4.3). Two sesquiterpene alcohols were also classed in this group and were highly correlated to one another. 00.20.40.60.811.21.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-bulnesene                (a)00.0050.010.0150.020.025can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)bulnesol                    (b)00.511.522.53can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)3-carene                     (c)95  00.10.20.30.40.50.60.70.80.9can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)p-cymene                    (d)00.10.20.30.40.50.60.7can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)p-cymenene                  (e)00.0050.010.0150.020.025can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-eudesmol               (f)00.050.10.150.20.250.3can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)cis-β-farnesene          (g)96    00.020.040.060.080.10.120.140.160.180.2can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-fenchene                  (h)024681012can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)linalool                       (i)00.511.522.533.544.5can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-phellandrene             (j)01234567can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)β-phellandrene           (k)97    00.020.040.060.080.10.12can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-santolene (l)00.020.040.060.080.10.12can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)δ-selinene                 (m)00.511.522.5can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-terpinene                  (n)00.20.40.60.811.21.41.61.8can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)γ-terpinene                 (o)98   Figure 4.6 Terpene profiles representing a unique group of terpenes that dominate both THC-dominant and CBD-THC hybrid strains. (a) α-bulnesene (b) bulnesol (c) 3-carene (d) p-cymene (e) p-cymenene (f) α-eudesmol (g) cis-β-farnesene (h) -fenchene (i) linalool (j) α-phellandrene (k) β-phellandrene (l) α-santolene (m) δ-selinene (n) α-terpinene (o) γ-terpinene (p) terpinen-4-ol (q) terpinolene (r) α-thujene. Error bars represent SEM, n=3.   The final 3 monoterpenes and 4 sesquiterpenes were predominantly found in CBD-containing strains (Figure 4.7). Two sesquiterpene alcohols, guaiol and 10-epi-γ-eudesmol were highly correlation to one another (Figure 4.3). 00.020.040.060.080.10.12can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)terpinen-4-ol              (p)051015202530354045can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)terpinolene                   (q)00.20.40.60.811.21.41.61.8can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)α-thujene                    (r)99 00.010.020.030.040.050.06can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)alloaromadendrene           (a)00.020.040.060.080.10.120.140.160.18can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)cis-α-bisabolene            (b)00.010.020.030.040.050.060.070.080.09can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)10-epi-γ-eudesmol           (c)00.010.020.030.040.050.060.070.08can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)guaiol                         (d)100   Figure 4.7 Terpenes predominantly found in higher CBD strains. (a) alloaromadendrene (b) cis-α-bisabolene (c) 10-epi-γ-eudesmol (d) guaiol (e) cis-β-ocimene (f) trans-β-ocimene (g) sabinene. Error bars represent SEM, n=3.  4.3.3 Aroma characterization Monoterpenes and sesquiterpenes are the volatile constituents responsible for the aromatic characteristics of Cannabis. The aromatic descriptors for each terpene identified in the strains were collected from published sources (Breitmaier, 2006) and are grouped according to their presence within the cannabinoid classes. (Table 4.2). The aromas range 0246810121416can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)cis-β-ocimene               (e)00.020.040.060.080.10.120.140.160.180.2can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)trans-β-ocimene             (f)00.20.40.60.811.21.4can34can38can30can35can19can16can14can39can41can31can28can40can23can33can12can22can10can20can37can42can11can25can13can27can36can24can21can29can15can18can32can17can26Peak Area (%)sabinene                    (g)101 from pine and woody to spicy, floral and citrus. While terpenes present in all or most strains are thought to invoke most of the major aromas, terpenes predominantly found in specific cannabinoid classes (as described in Section 4.3.2) are considered undertones contributing to the unique aromas within each cannabinoid class. The terpenes associated with mid-level THC-dominant strains were a combination of woody, floral and herbal undertones. The terpenes found in THC-dominant strains contained herbal, floral, woody, sweet and spicy undertones. The unique cluster of terpenes correlated with terpinolene found in THC-dominant and THC-CBD hybrid strains have a considerable number of citrus, woody, musty, floral and sweet descriptors. The CBD-containing strains have primarily citrus, tropical and sweet undertones.   102 Table 4.2 Aroma descriptors for each of the terpenes identified within the Cannabis strains, grouped based on their presence within the cannabinoid classes. Group Terpene Scent Descriptors Aroma All or most Strains -pinene woody/pine woody, pine, citrus, spicy, floral β-pinene woody/pine trans-2-pinanol pine camphene woody/camphor -gurjunene derivative woody/balsamic β-maaliene woody selina-3,7(11)-diene possible woody camphene hydrate woody/camphor -guaiene woody 4,11-selinadiene woody endo-borneol camphor fenchone camphor Z-sabinene hydrate balsam -gurjunene musty β-myrcene spicy/balsamic/peppery caryophyllene spicy/cloves/roses copaene spicy/honey -muurolene spicy limonene citrus -terpineol citrus β-cubebene citrus valencene citrus guaia-3,9-diene floral, rose, geranium germacrene A floral ylangene ylang ylang humulene hoppy -selinene celery exo-fenchol basil β-selinene celery Mid-THC strains -gurjunene woody/balsamic floral, woody, herbal santolina triene floral sesquiterp-1 n/a -cadinene herbal/thyme THC-Dominant strains -amorphene woody herbal, woody, floral, citrus caryophyllene oxide spicy, woody/carrot germacrene B floral/roses -elemene floral 2-carene sweet (Z,Z)--farnesene citrus -cubenene herbal β-elemene herbal β-sesquiphellandrene herbal/oregano    103 Group Terpene Scent Descriptors Aroma High THC or THC/CBD  strains -thujene woody/frankincense citrus, pine, woody, sweet, spicy -terpinene woody -santalene woody -bulnesene patchouli -fenchene camphor cis-β-farnesene citrus/sweet p-cymene citrus/sweet -phellandrene citrus/pepper -terpinene citrus terpinolene sweet/pine/citrus linalool citrus/floral/sweet -selinene floral 3-carene sweet -eudesmol sweet terpinen-4-ol peppery/musty/sweet p-cymenene spicy/cloves β-phellandrene minty bulnesol spicy CBD containing strains alloaromadendrene woody citrus, woody, sweet, tropical guaiol rose wood 10-epi--eudesmol sweet cis--bisabolene citrus/myrrh/balsamic cis-β-ocimene citrus/tropical trans-β-ocimene citrus/tropical sabinene citrus/pine/spicy  4.3.4 Terpene Metabolomics A principal component analysis (PCA) was performed on the autoscaled terpene profiles to evaluate the clustering and multivariate correlations between the metabolites. The PCA is shown in Figure 4.8. The first two PCs describe 47.53 % of the variance within the data. There is no distinct clustering of the strains according to their THC/CBD classifications as all five cluster groups overlap significantly. Based on the loadings plot of the first two PCs (Figure 4.9), the majority of the sesquiterpenes cluster together in the top right quadrant of the plot, while the terpinolene-correlated monoterpenes appear to cluster separately from the remaining strains. PC2 appears to have some influence by different monoterpenes: α-pinene and β-myrcene are negatively correlated from the terpinolene-correlated terpenes on this PC. 104  Figure 4.8 Principal component analysis (PCA) of the monoterpene and sesquiterpene profiles for the Cannabis dataset. Strains were prepared and analyzed in triplicate. They are color coded according to their THC/CBD contents described in Chapter 3.  Figure 4.9 PCA loadings plot of PC 1 and PC2 describing the influence of terpenes on the variation of Cannabis strains in Figure 4.8. -10 -5 0 5 10-8-6-4-20246810Scores on PC 1 (33.62%)Scores on PC 2 (13.81%)105  It was previously noted that many of the monoterpenes and sesquiterpenes were identified across different cannabinoid classes. Therefore, a data reduction strategy was undertaken to remove these metabolites and identify any unique clustering of the strains and metabolites when removing these terpenes. In this case, the number of metabolites was reduced from 67 to 38 and then subjected to principal component analysis. Figure 4.10 illustrates the first two PCs of this reduced dataset, where 40.02% of the data is described in the first two PCs. Figure 4.11 describes the loadings plot for the two PCs. The first PC is strongly influenced by the terpinolene-correlated monoterpenes, for which the strains all cluster together on the right side of the scores plot. PC2 appears to cluster a few strains on the top left and bottom left quadrant from the majority of the remaining strains. These are influenced by the contents of several sesquiterpenes. The strains in the top left quadrant are impacted by δ-selinene, germacrene B, α-cubenene and γ-elemene, all metabolites identified to be present only in THC-dominant strains.   Figure 4.10 PCA of the monoterpene and sesquiterpene profiles within the Cannabis dataset after implementing a data reduction strategy. Strains were prepared and analyzed in triplicate. They are color coded according to their THC/CBD contents described in Chapter 3.  106  Figure 4.11 PCA loadings plot for monoterpenes and sesquiterpenes after implementation of the data reduction strategy to identify the metabolites influencing the strain clustering of the PCA scores plot in Figure 4.10.  4.4 Discussion The data indicate that the majority of the terpenes are common across most strains but that some differences exist that can be associated with the major cannabinoids. As aromas are key contributors to the perceptions of potency and selection, this provides a significant outlook on domestication as it may have impacted the breeding selections for different strains (Gilbert & DiVerdi, 2018). There has been anecdotal evidence suggesting that clandestine breeders can predict the potency of THC strains based on slight aromatic undertones and breeders selecting for CBD-containing strains are also selecting for specific aromas to predict these metabolites (Clark & Merlin, 2013). Within the dataset there were several terpenes which were more prominent within a specific cannabinoid class, confirming selection pressures based on breeders selecting those strains. For the high THC strains, there was a higher prevalence of herbal and floral undertones with higher prevalence of several 107 sesquiterpenes. Caryophyllene oxide was one of the sesquiterpenes found in high THC strains, which is the aromatic compound used by sniffing dogs to detect drugs (Mediavilla et al., 1997). The CBD-containing strains were higher in citrus and tropical undertones which were attributed to several monoterpenes, and sesquiterpene alcohols. Aromas are determined based on volatility, threshold, concentration and interactions with other aromatic compounds, therefore the data described are a preliminary estimation of aromatic characteristics from each compound (Wu et al., 2016). Headspace GC-MS analysis was employed for profiling monoterpenes and sesquiterpenes in Cannabis because of its sensitivity in comparison to solvent extraction methods and ability to highlight the aromatic expression (headspace) of the strains. A total of 67 different metabolites were identified against reference standards and the NIST spectral database with considerable matching capabilities. While a few of the sesquiterpenes could not be identified, by having a large database of samples, it was possible to select strains with the highest prevalence of different terpenes to increase spectral matching. In many previous characterizations of Cannabis, the number of terpenes ranged from 14 to 37, focusing only on high abundance terpenes (Elzinga et al., 2015; Hazekamp & Fischedick, 2012; Fischedick et al., 2010; Hazekamp et al., 2016; Jin et al., 2017). Over 120 different terpenes have previously been detected in Cannabis, but many of those not detected are typically present in trace levels (Hazekamp et al., 2016). The implementation of this more sensitive technique due to the detection of more low concentration terpenes provides a deeper insight into the phytochemical variation within strains and the underlying variances that would otherwise be overlooked in traditional solvent extraction-based methods. Headspace has the potential to represent volatile constituents that may have a stronger influence on aroma due to their higher volatility at room temperature. Although quantitation is not easily interpreted with this data, the comparisons of relative percentages within the monoterpene class and within the 108 sesquiterpene class provides valuable insight into relative proportions and presence/absence of those metabolites. The volatility of sesquiterpenes is lower in comparison to monoterpenes, therefore it is not possible to compare the relative intensities of these two classes of metabolites. Autoscaling the data prior to metabolomic evaluation overcomes these issues by scaling each metabolite to unit variance, therefore having equal weight in the PCA model.  The increased detection of the volatile metabolites has allowed evaluation of minor terpenes that would otherwise not be detected with solvent extraction methods. There are no losses from solvent removal or concentration steps to improve sensitivity as the samples are ground and immediately placed in gas tight headspace vials to ensure minimal losses after breaking the trichomes and releasing the metabolites. While previous research on major terpenes has provided limited information on metabolite correlations, this data has provided considerable correlations between terpenes such as terpinolene and 9 other low abundance monoterpenes that have not previously been identified. The abundance of a metabolite does not necessarily correlate with pharmacological significance, therefore understanding the underlying variation of Cannabis could help substantiate the many anecdotal claims of Cannabis pharmacology and patient use (Russo, 2011; McPartland & Russo, 2001).  Several high abundance sesquiterpenes were present in all Cannabis strains but are not commonly evaluated when profiling Cannabis strains: β-maaliene, guaia-3,9-diene and selina-3,7(11)-diene. These sesquiterpenes were present in proportions as high as humulene in many of the strains. Their presence in all strains indicates they have not been selected or bred out any strains within the dataset evaluated and may have pharmacological significance. For example, β-maaliene was isolated from Nardostachys chinensis and reduced locomotor activity after inhalation due to the sedative effect of this compound (Takemoto et al., 2009). Sedation is a common effect noted for many “indica” strains. Guaia-3,9-diene is prevalent in many different plants including Atractylodes spp., Curcuma wenyujin, Blumea balsamitera, 109 Eucalyptus spp. and Piper longum L., but has not been studied as a single entity. Extracts of these other plants have activities including immune boosting, digestive aid, anti-inflammation, rheumatism healing and many more (Peng et al., 2011; Pang et al., 2014; Al-Snafi, 2017; Hieu et al., 2018).  Selina-3,7(11)-diene (also known as eudesma-3,7(11)-diene) was detected in considerable levels in Brazilian green propolis commonly used in folk medicine to fight infections but has also not been evaluated as a single entity (de Albuquerque et al, 2008). Monoterpenes and sesquiterpenes are found in many plants. There are over 30000 known terpenes that exist in nature (Breitmaier, 2006; Tholl 2015). With Cannabis there is considerable emphasis on the “entourage effect” theory that considers the effects of different cannabinoids and/or terpenes present in the strains based on their known pharmacological effects (Russo, 2011; McPartland & Russo, 2001). A summary of the pharmacological effects of the terpenes identified in Cannabis through in vitro and in vivo studies are summarized in Table 4.3. Only those which have been studied as single entities are summarized. No activities were available for terpenes identified in group 2. Major monoterpenes such as α-pinene, β-myrcene and limonene have been shown to have anti-inflammatory, analgesic and sedative properties evaluated in animal models, respectively (Yun, 2014; Gurgel do Vale et al., 2002; Kim et al., 2015; Freitas et al., 1993). With these metabolites present in all strains, these properties will depend on relative abundance and possible synergistic, additive or negative interactions with other metabolites. Terpinolene, present in high abundance in only a select few strains, also has anti-inflammatory and sedative properties in animal models (Macedo et al., 2016; Ito & Ito, 2013).    110 Table 4.3 Pharmacological activities of monoterpenes and sesquiterpenes identified in Cannabis. Group Terpene Pharmacological Activity References Group 1 camphene expectorant Boyd & Sheppard, 1970 caryophyllene anti-inflammatory, antinociceptive, anxiolytic, antispasmodic, antidepressant, gastroprotective Galdino et al., 2012; Katsuyama et al., 2013; Fernandes et al., 2007; Leonhardt et al., 2010; Bahi et al., 2014; Cho et al., 2007 fenchone antinociceptive activty Him et al., 2008  humulene anti-inflammatory, anti-tumor Fernandes et al., 2007; El Hadri et al., 2010 limonene antioxidant, anticancer, sedative Yun, 2014; Gurgel do Vale et al., 2002; Miller et al., 2013; Murali et al., 2013 β-maaliene sedative Takemoto et al., 2009 β-myrcene sedative, analgesic, anti-oxidant, Freitas et al., 1993; Paumgartten et al., 1990; Cifti et al., 2011 α-pinene antinociceptive activty, anti-inflammatory, anxiolytic Him et al., 2008; Kim et al., 2015; Satou et al., 2014 β-pinene Anti-depressant Guzman-Gutierrez et al., 2015 α-terpineol anti-inflammatory, antinociceptive, gastroprotective de Oliveira et al., 2012; Souza et al., 2011 valencene anti-cancer, anti-inflammatory Nam et al., 2016; Tsoyi et al., 2011 Group 3 caryophyllene oxide analgesic, anti-inflammatory, anti-cancer Chavan et al., 2010; Park et al., 2011 β-elemene anti-tumor Chen et al., 2011 β-sesquiphellandrene anti-cancer Tyagi et al., 2015 Group 4 α-bulnesene anti-platelet Hsu et al., 2006 p-cymene antinociceptive, anti-inflammatory, antioxidant Quintans et al., 2013; de Oliveira et al., 2015. β-Eudesmol anti-inflammatory, muscle relaxant, anti-cancer, appetite stimulation, anti-angiogenic, gastroprotective, anticonvulsant Seo et al., 2011; Kimura et al., 1991; Plengsuriyakarn et al., 2015; Ohara et al., 2017; Kimura & Sumiyoshi, 2012; Choiu et al., 1997 linalool Antidepressant, antinociceptive, sedative Guzman-Gutierrez et al., 2015; Peana et al., 2003; Kuroda et al., 2005 α-phellandrene antinociceptive Lima et al., 2012 γ-terpinene antioxidant, antinociceptive Milde et al., 2004; Passo et al., 2015  111 Group Terpene Pharmacological Activity References Group 4 terpinen-4-ol antimicrobial, antihypertensive, anticonvulsant, anti-cancer Giwanon et al., 2000; Lahlou et al., 2003; de Sousa et al., 2009; Calcabrini et al., 2004 terpinolene antinociceptive, anti-inflammatory, sedative Macedo et al., 2016; Ito & Ito, 2013 Group 5 alloaromadendrene antioxidant, Yu et al., 2014 cis-α-bisabolene anticonvulsant Orellana-Paucar et al., 2012 guaiol anti-inflammatory Ringrose et al., 1975 sabinene antimicrobial Giwanon et al., 2000  There is considerable anecdotal evidence that Cannabis strains with similar THC/CBD contents (or those within a cannabinoid class) exhibit varying pharmacological effects (Russo, 2011; Hazekamp et al., 2016). This is likely due to underlying phytochemical diversity across the strains. In several cases, monoterpenes and sesquiterpenes were identified within a single cannabinoid class, but not present in all the strains. This provides evidence that there may be impacts from these low concentration terpenes. For example, α-thujene, cymene, phellandrene, linalool and several others were all identified in group 4 terpenes. Linalool has been shown to have anti-inflammatory, sedative, anxiolytic, anti-convulsant and anti-depressant activities (Linck et al., 2010; Russo, 2011). Cymene has antinociceptive activity (Bonjardim et al., 2012). Terpinen-4-ol has been studied extensively for its anticonvulsant and anti-cancer activities (de Sousa et al., 2009; Calcabrini et al., 2004).  The activities of many sesquiterpenes vary considerably. The two major sesquiterpenes: β-caryophyllene and humulene have anti-inflammatory properties, while β-caryophyllene has been shown to have anti-cancer, anxiolytic and anti-depressant activities (Bahi et al., 2014; Rogerio et al., 2009; Russo, 2011). Some sesquiterpenes of note are β-elemene which has a strong anti-cancer activity, eudesmol has antiangiogenic properties and bisabolene has anti-convulsant activity (Chen et al., 2011; Tsuneki et al., 2005; Orellana-112 Paucer et al, 2012). Bisabolene was identified to have a higher prevalence with CBD-containing strains and while there is considerable evidence that CBD has seizure-reduction activity (Devinsky et al., 2014), this correlated terpene could signify an additive or synergistic effect, supporting the benefits of whole plant efficacy versus isolated cannabinoids. This further substantiates the need for in-depth metabolomic evaluation of strains for pre-clinical and clinical studies and the addition of metabolomic and chemometric analysis to investigate these relationships and possible hypotheses.  The pharmacological effects of many of the terpenes in Cannabis have not been studied extensively, but many plant essential oils with similar terpene compositions to Cannabis have been evaluated. For example, Salvia and Ocimum santum (holy basil) are used for their analgesic, anti-depressant, anxiolytic and anti-inflammatory activities (Garg & Sardana, 2016; Fu et al., 2013). These plants have many similar terpenes including borneol, β-pinene, α-pinene, camphene, α-thujene, β-caryophyllene, sabinene, limonene, p-cymene, terpiniolene, ocimene, α-cubebene, linalool, β-elemene, β-caryophyllene, α-guaiene, α-amorphene, humulene, isoborneol, α-selinene, β-selinene, and α-muurolene. Myrcia spp. have many similar terpenes and exhibit anti-inflammatory, antiproliferative and anti-nociceptive activities (Cascaes et al., 2015). Another similar plant is Ocium basiclicum which can impact the central nervous system with anti-depressant and anti-convulsant activities (Bariyah et al., 2012). Unfortunately, there is still a considerable amount of research needed to understand the underlying impacts of these metabolites and their pharmacological significance in humans. Amino acid sequences coding for terpene synthases are subdivided into six TPS subfamilies from TPSa to TPSf (Bohlmann et al., 1998). Monoterpenes are products of terpene synthases (TPS), derived from TPSb genes, using geranyl pyrophosphate as a precursor. Recently, several terpene synthases have been identified in Cannabis, many of 113 which are responsible for the production of the major monoterpenes: limonene, α-pinene, β-myrcene and a few lower abundance monoterpenes such as α-terpinene, γ-terpinene, and Z/E-β-ocimene (Booth et al., 2017). With multiple TPSb enzymes producing the same monoterpene, breeding selection would need to breed out multiple TPS genes to remove one of the major terpenes. An exception to this is terpinolene, which is either present in very low amounts, or it is one of the predominant monoterpenes in the strains. To date, the TPS responsible for terpinolene was not identified (Booth et al., 2017). Terpinolene is a strong pine/citrus aroma that could be one selection pressure for the six high terpinolene strains, which represents a clear break in the biosynthetic pathway. TPSb enzymes produce only a few of the major monoterpenes and it has yet to be determined how the additional low concentration monoterpenes are produced. Post-translational modifications may be occurring in the trichomes during growth, processing and handling impacting to production of these low concentration terpenes (Kersten et al., 2015; Tholl, 2015). Sesquiterpenes are products of terpene synthases, derived from TPSa genes, using farnesyl pyrophosphate as a precursor. Four TPSa genes have been characterized in Cannabis (Booth et al., 2017). One TPSa enzyme is responsible for the production of β-caryophyllene and humulene, while the other three appear to produce several different sesquiterpenes. For example, CsTPS7FN produced 23 different sesquiterpene metabolites (Booth et al., 2017). As the prevalence of β-caryophyllene and humulene is ubiquitous across the Cannabis dataset and the ratio of the two is relatively consistent between strains, it is evident that breeding has not impacted the synthesis of these metabolites. Also, many of the sesquiterpenes appear to be weakly correlated with one another based on the Pearson correlations and principal component analyses. As sesquiterpenes are less volatile in comparison with monoterpenes, there may not have been pressure to select different sesquiterpenes when assessing by aroma. Additionally, as a single TPS can produce many 114 sesquiterpenes, breeding out a specific TPS gene would lead to the loss of many sesquiterpenes. For example, strains can19, can28 and can41 are all lacking a group of sesquiterpenes that appear to be present in almost all other strains: 4,11-selinadiene, β-selinene, γ-gurjunene, α-selinene and α-bulnesene. While a TPSa gene has not been identified for these sesquiterpenes, it may be possible that there is a TPS present in most strains that is not expressed in these three strains. As these strains are present in different classes (orange – high THC/high CBD and purple – high THC), selection may not just depend on major cannabinoids, where other relationships within the terpene clusters themselves are occurring. Unsupervised clustering of the data is useful to visualize the major correlations within the dataset. These methods do not consider the original classification of the strains and can provide valuable information on the clustering of the samples. PCA highlighted the clustering of the strains based solely on the terpene composition without the impact of cannabinoids. Using the entire terpene dataset, there is clear clustering of the terpinolene-dominant strains. These strains have a considerably different monoterpene profile in comparison to the other strains. There is limited distinction separating the cannabinoid classes with this model. By implementing data reduction strategies and modeling only the terpenes that were identified in unique cannabinoid classes within the dataset, more structure in the data can be visualized and clustered to get a better understanding of the impacts of different terpenes. For example, can26 clusters with can12 on the top left of the scores plot separately based on the presence of four sesquiterpenes (δ-selinene, germacrene B, α-cubenbene and γ-elemene). These strains, while classified as very high THC strains (blue), they have a distinct composition relative to the other strains. Can26 is outside of the 95% confidence level indicating that this strain is varies considerably compared to the entire dataset. 115 Cannabis aroma plays many roles in strain selection, euphoria, product quality and is strongly associated with clandestine breeding (Clarke & Merlin, 2013; Gilbert & DiVerdi, 2018). Many of the terpenes have similar characteristic aromas, which can be impacted by concentration, synergy with other aromatic compounds and subjective interpretation of aroma (Wu et al., 2016). With Cannabis, subjective interpretation of aroma during breeding can impact terpene profiles. Several dominant aromas can be described in Cannabis strain names including lemon, sour, skunk, berry/fruit, diesel, or cheese. To date, it has not been possible to definitively correlate the specific aromas to the terpenes. There has been considerable variation observed between strain name and chemical composition, suggesting that some strain names may not accurately describe aroma (Elzinga et al., 2015). Based on discussions with a Cannabis producer/breeder, producers will “pop” several seeds of a specific strain to select the characteristics they are expecting which will impact the underlying phytochemical composition and diversity of Cannabis.  These findings present a further indication of the impacts of breeding closely related plants and its impacts on secondary metabolism leading to domestication syndrome in the Cannabis genome (Meyer et al., 2012). In many of the strains there is a loss of phytochemical diversity due to the presence of terpenes found only in specific cannabinoid classes and further emphasized by the clear breaks in expression of the terpinolene and its correlated monoterpenes in Cannabis strains. Breeding selection will impact the fitness of this genus as these terpenes may be responsible for enhancing pest resistance, improving pollination or other natural survival mechanisms (Tholl, 2006). Based on these data, the impacts of human preferences and selections appears to have reduced the phytochemical diversity. The proposed “entourage effect” used to describe the variation in pharmacological effects of medicinal Cannabis is thought to be related to synergistic and additives effects between cannabinoids and other secondary metabolites, including terpenes (Russo, 2011). With the 116 loss of phytochemical diversity between strains, this will further impact the pharmacological effects of the strains and potentially the loss of some medicinal effects that are desirable to patients. Domestication in Cannabis has led to selection of germplasm that is genetically similar but with a high degree of phytochemical diversity. This led to vastly different experiences and efficacy for recreational and medicinal users, respectively, that can ultimately impact the future of medicinal Cannabis use and acceptance by medical professionals. Future research is needed to understand the pharmacology of low abundance terpenes and synergistic effects in Cannabis strains, to determine importance for medical efficacy and their roles in plant biosynthesis. 117 Chapter 5: Multiblock Data Fusion Model to Identify Relationships Between Cannabinoids and Terpenes 5.1 Synopsis Both cannabinoids and terpenes are produced in the glandular trichomes of the female inflorescence of Cannabis (Turner et al., 1980). Cannabinoids are biosynthesized by the condensation of polyketides such as olivetolic acid or divarinolic acid with geranyl pyrophosphate (GPP). GPP is a common precursor used in cannabinoid and monoterpene biosynthesis pathways which is produced through the MEP pathway as shown in Figure 5.1 (Flores-Sanchez & Verpoorte, 2008). While the monoterpene and cannabinoid pathways compete for the same precursor, sesquiterpenes are produced from FPP primarily through the MVA pathway. Cannabinoid biosynthesis also relies on production of polyketides through the polyketide synthase enzyme (Degenhardt et al., 2017). The production of precursors such as CBGA and CBGVA are produced through geranyl diphosphate:olivetolate geranyltransferase (GOT) and then with THCA and/or CBDA synthase, THCA and CBDA products are formed. Monoterpenes and sesquiterpenes are produced from their precursors GPP and FPP, respectively, through several terpene synthases that have previously been identified in Cannabis (Booth et al., 2017). The cannabinoid and terpene profiles vary depending on the synthases expressed within the different strains. In addition to purported pharmacological links, the close proximity of the biosynthesis and accumulation of the classes of metabolites provides further opportunities for there to be unique biosynthetic relationships.  118  Figure 5.1 Biosynthetic pathway of geranyl pyrophospate via the MEP pathway and olivetolic acid via the polyketide synthesis pathway to produce the precursors for monoterpene and cannabinoid biosynthesis. Enzymes are bolded in orange text. Abbreviations: DXS – Deoxy-D-xylulose phosphate (DXP) synthase, DXR – DXP reductase, MCT – 4-diphosphocytidyl-2-C-methyl-D-erythritol synthase, CMK - 4-diphosphocytidyl-2-C-methyl-D-erythritol kinase, MDS – 2-C-methyl-D-erythritol 2,4-cyclodiphospate synthase, HDS – 4-hydroxy-3-methylbut-2-enyl diphosphate synthase, HDR - 4-hydroxy-3-methylbut-2-enyl diphosphate reductase, GPS – geranylpyrophospate synthase, TKS – tetraketide synthase, OAC – olivetolic acid cyclase, GOT – geranylpyrophosphate:olivetolate geranyl transferase. THCAS – THCA synthase, CBDAS – CBDA synthase, CBCAS – CBCA synthase.  pyruvate glyceraldehyde-3-phosphateMEP PathwayDXSDXR1-deoxy-D-xylulose 5-phosphate2-C-methyl-D-erythritol 4-phosphateMCTCMKMDSHDS4-diphosphocytidyl-C-methyl-D-erythritol4-diphosphocytidyl-C-methyl-D-erythritol 2-phosphate2-C-methyl-D-erythritol 2,4-cyclodiphosphate4-hydroxy-3-methylbut-2-enyl diphosphateIsopentenyl diphosphate (IPP)Dimethylallyl diphosphate (DMAPP)GPSHDRGeranyl pyrophosphate (GPP)Monoterpene synthase(s)MonoterpenesTKS +2 malonyl CoAhexanoyl CoAOACTKS + malonyl CoApentyl tetra-β-ketideCoApentyl tri-β-ketideCoAolivetolic acidGOTCBGATHCAS/CBDAS/CBCAScannabinoidsPolyketide Synthesis119 The application of chemometric models to large metabolomic data sets provides a platform to study underlying relationships within the data (Turi et al., 2015; Hall, 2011). The exploration of data through unsupervised models provides unbiased discovery of relationships between metabolites and samples, however, incorporating data from multiple instruments or platforms can be challenging (Turi et al., 2015; Tugizimana et al., 2013). Data fusion models are one technique to overcome issues as they treat each data set as a different block and scale each block to have equal variance (Boccard & Rudaz, 2014; Abdi et al., 2013). It was hypothesized that use of a data fusion model could elucidate underlying relationships between cannabinoids and terpenes and lead to the discovery of unique biosynthetic pathways in some strains. To evaluate this hypothesis, the cannabinoid and terpene data of the thirty-three Cannabis strains were subjected to multi-block data fusion models to explore the underlying relationships between the three metabolite classes: cannabinoids, monoterpenes and sesquiterpenes. The lot-to-lot comparisons of nine strains was undertaken to verify the findings. 5.2 Experimental 5.2.1 Plant materials The thirty-three Cannabis strains purchased from five licensed producers described in Chapter 2 were used for this work. New lots for nine strains (can15, can17, can18, can19, can21, can22, can24, can25 and can26) were purchased from the same licensed producers approximately 18-20 months after the original lots were purchased in order to evaluate lot variance. Samples were provided as 5, 10 or 15 g samples as dried whole flowers or milled in sealed containers. 120 5.2.2 Metabolite profiling Original data from cannabinoid quantitation and terpene headspace profiling were used in this chapter. New lots of the same strains were analyzed for cannabinoid content and terpene profiling according to methods described in sections 3.2.3 and 4.2.2, respectively. Chromatograms were manually integrated and quantified as described in the previous sections.  5.2.3 High-level data fusion Data fusion models were used to evaluate the original datasets by separating cannabinoids, monoterpenes and sesquiterpenes into separate data blocks. Each block was analyzed by principal component analysis and the first five components for each block was fused for the final data modeling. Multiple Factor Analysis (MFA) was undertaken in R using the package FactoMineR and clustering using hierarchical clustering analysis (HCA) using Ward’s agglomerative algorithm were undertaken in SOLO+MIA. 5.2.4 Chemometrics Pearson correlations were calculated in R. Clustering algorithms (Clustend & Hopkins statistic) were evaluated in R to determine the clustering tendency and optimal number of clusters in the HCA output. Strains reclassified based on their terpene and cannabinoid compositions were evaluated with one-dimensional self-organizing maps (1DSOM) and SAM statistic to identify metabolite clusters and those impacting different classes (Meinicke et al., 2008). These data were used to develop the decision tree (rpart) for classifying strains according to their chemical compositions. Bar graphs comparing cannabinoid profiles and terpene profiles were generated to compare the two lots of the same strains. Ratios of THC to CBD content were determined by calculating total THC and total CBD as described previously and the ratio was calculated by 121 total THC divided by total CBD. Samples with undetected levels of CBD were deemed to have THC:CBD ratios designated as >100. Variance was calculated for each metabolite between the different lots of the strains. The total variance was calculated by summing the variance of each metabolite. Variance of the minor metabolites was determined as the total variance excluding the variance for CBDA, CBD, THCA and THC. 5.3 Results 5.3.1 Interrelationships between terpenes and cannabinoids Pearson correlations were calculated for the fused cannabinoid-terpene dataset prior to high-level data fusion to observe relationships and correlations between these metabolite classes. While many of the cannabinoids and terpenes are not strongly correlated, a few correlations were observed (Figure 5.2). CMPD20 is strongly correlated with 3 terpenes: santolina triene (0.80), sesquiterp-1 (0.99) and δ-cadinene (0.99). CMPD7, CMPD10, CBGA and CMPD12 are correlated with terpinolene, with CMPD12 being having the strongest correlation of 0.75, and the other three with correlation coefficients between 0.42 and 0.64. CMPD10 is also strongly correlated with cis-β-ocimene with a p = 0.88. CMPD3 and CMPD4 both have weak positive correlations with different sesquiterpenes, but CMPD3 is strongly correlated with linalool (0.80). 122  Figure 5.2 Pearson correlation heatmap describing correlations between cannabinoids and terpenes.  5.3.2 High-level data fusion of terpenes and cannabinoids The underlying relationships between cannabinoids, monoterpenes and sesquiterpenes were evaluated using the multiblock data fusion model, multiple factor analysis (MFA). Each metabolite class was separated into individual blocks and modeled separately, of which the first five components from each of the metabolite blocks were used to construct the data fusion model. The model was generated on the entire data set and the factor map describing the clustering of the different strains is shown in Figure 5.3. In the first two dimensions of the model, 38.6% of the variance in the data is captured. In the first dimension, the majority of the variation in the data and clustering is based on THC/CBD contents as described in the correlation plot (Figure 5.4). The second dimension is considerably impacted by the terpene profiles of the strains. 123  Figure 5.3 Unsupervised data fusion clustering with a multiple factor analysis (MFA) factor plot describing dimensions 1 and 2 of the cannabinoid, monoterpene and sesquiterpene profiles within the Cannabis dataset. Strains are color coded according to their cannabinoid classes (n=3).   Figure 5.4 Multiple factor analysis correlation plot of the cannabinoids, monoterpenes and sesquiterpenes within the Cannabis dataset. 124  The top left quadrant of Figure 5.3 describes the clustering of the terpinolene strains, which are impacted by the separation in the second dimension of the plot. The CBD-containing strains separate from the THC-dominant strains on the first dimension. The correlation plot in Figure 5.4 indicates correlation between many of the terpinolene-correlated terpenes and two unknown cannabinoids (CMPD10 and CMPD12). CBGA also appears to be correlated with these metabolites.  Within the terpenes, there is correlation between the sesquiterpene alcohols and CBDA, which was previously noted in Chapter 4. THCA does not strongly correlate with any metabolites. 5.3.3 Metabolomic based chemotyping of Cannabis strains To develop a preliminary decision tree-based classification of the cannabinoid and terpene data, the clustering tendency of the data was evaluated. A Hopkins statistic of 0.08 indicates a clear clustering tendency. The data were then clustered with agglomerative hierarchical clustering analysis (HCA). The clusters are described in the HCA dendrogram in Figure 5.5 for which nine clusters were identified according to the NbClust algorithm and composed of 1 to 7 different strains. For determining the metabolites responsible for class differentiation several techniques were employed: SAM statistic, 1D-SOM and subtractive metabolomics. The data from these models were used to develop the decision tree for class determination. 125  Figure 5.5 Hierarchical cluster analysis of the strains based on their cannabinoid and terpene compositions. The samples are colored according to their cannabinoid class.  The SAM statistic identified 73 out of 99 metabolites as significant with a FDR of 0.001 (Table 5.1). The 1D-SOM was evaluated for nine metabolite clusters, which are summarized in Table 5.2. The significance of each cluster relative to the respective class is summarized in the table as “strongest class correlation”. Subtractive metabolomics was employed for identification of unique metabolites within each class. In this case, class 3 (CAN23) and class 7 (CAN14) were found to have unique profiles. CAN14 was the only strain with CMPD-3, while CMPD-20, the unidentified terpene labeled sesquiterp-1 and δ-cadinene were unique to CAN23. Based on the statistical models identifying metabolite clusters and significance, a decision tree was developed to classify strains (Figure 5.6). The decision tree was developed with eight metabolites including terpinolene, linalool and CBDA. In the case of classes 4 and 9, it was necessary to have multiple nodes resulting in the same class due to the variance observed within the classes. 126 Table 5.1 Results of the significance analysis of microarray to identify the cannabinoids and terpenes significant in the chemotaxonomic classification of Cannabis strains. Metabolite d.value stdev rawp q.value sesquiterp-1 12488.04 1.32E+07 0 0 δ-cadinene 3433.98 5.05E+06 0 0 β-phellandrene 236.83 3.77E+13 0 0 p-cymene 159.73 5.24E+11 0 0 3-carene 149.99 1.06E+13 0 0 α-thujene 129.06 3.61E+12 0 0 α-phellandrene 113.44 2.85E+13 0 0 α-terpinene 86.96 1.20E+13 0 0 terpinolene 76.79 4.23E+15 0 0 β-cubebene 57.56 2.90E+10 0 0 γ-terpinene 56.94 7.82E+12 0 0 valencene 55.13 9.65E+10 0 0 selina-3,7(11)-diene 54.63 3.39E+12 0 0 α-gurjunene derivative 53.84 5.80E+09 0 0 guaia-3,9-diene 51.45 1.49E+12 0 0 CMPD12 50.8 6.77E-05 0 0 santolina triene 48.57 1.21E+11 0 0 CMPD1 46.17 6.59E-04 0 0 CMPD18 45.23 1.39E-04 0 0 germacrene A 44.95 5.03E+10 0 0 CBDA 41.06 5.09E+00 0 0 CMPD17 38.27 6.63E-04 0 0 β-maaliene 37.04 2.23E+11 0 0 γ-muurolene 34.38 8.57E+09 0 0 CMPD16 28.85 1.63E-03 0 0 terpinen-4-ol 28.77 4.36E+10 0 0 CBDVA 28.63 3.30E-04 0 0 THCA 27.89 1.04E+01 0 0 trans-2-pinanol 26.7 2.17E+12 0 0 δ-selinene 25.94 1.39E+10 0 0 β-caryophyllene 24.45 5.69E+13 0 0 copaene 24.3 7.77E+08 0 0 exo-fenchol 23.98 9.77E+12 0 0 endo-borneol 22.46 9.19E+10 0 0 ylangene 21.63 1.44E+10 0 0 camphene hydrate 20.96 9.50E+09 0 0 sabinene 19.68 1.52E+12 0 0 humulene 19.17 4.48E+12 0 0 4,11-selinadiene 18.44 2.18E+10 0 0 β-selinene 17.75 2.22E+11 0 0 127 Metabolite d.value stdev rawp q.value p-cymenene 17.67 2.03E+12 0 0 CBD 14.98 1.96E-02 0 0 α-fenchene 14.72 1.27E+11 0 0 CBGA 12.67 1.53E-01 0 0 CMPD6 12.31 1.79E-05 0 0 α-selinene 12.06 2.38E+11 0 0 CMPD11 10.08 6.63E-04 0.000101 3.01E-05 cis-α-bisabolene 9.96 3.51E+10 0.000101 3.01E-05 cis-β-farnesene 9.95 2.90E+11 0.000101 3.01E-05 limonene 9.57 1.39E+16 0.000101 3.01E-05 α-cubenene 9.42 3.72E+08 0.000101 3.01E-05 α-terpineol 9.37 3.66E+11 0.000101 3.01E-05 α-guaiene 9.26 4.03E+12 0.000101 3.01E-05 CBG 8.96 1.89E-03 0.000101 3.01E-05 cis-β-ocimene 8.22 1.09E+15 0.000101 3.01E-05 germacrene B 8.18 3.61E+11 0.000101 3.01E-05 2-carene 7.93 5.30E+10 0.000202 5.91E-05 γ-gurjunene 7.23 1.97E+10 0.000303 8.43E-05 CMPD7 6.97 2.40E-03 0.000303 8.43E-05 CMPD3 6.97 1.77E-08 0.000303 8.43E-05 CMPD10 6.81 2.03E-03 0.000505 0.000138 fenchone 6.58 1.17E+12 0.000606 0.000163 alloaromadendrene 6.27 3.17E+09 0.000707 0.000187 camphene 5.59 7.71E+13 0.00101 0.000263 caryophyllene oxide 5.4 9.99E+08 0.001111 0.000285 10-epi-γ-Eudesmol 5.37 9.97E+09 0.001313 0.000327 guaiol 5.3 6.47E+09 0.001313 0.000327 α-gurjunene 5.04 2.63E+09 0.001818 0.000446 β-myrcene 4.99 3.02E+16 0.00202 0.000488 THCV 4.93 3.06E-02 0.002121 0.000506 linalool 4.55 1.25E+14 0.002727 0.000641 THC 4.32 1.91E-01 0.004141 0.00096 CMPD20 4.27 2.55E-07 0.004242 0.000969    128 Table 5.2 Results of the 1D-self organizing map (1DSOM) for identifying the metabolite clusters responsible for the chemotaxonomic classification of Cannabis. Metabolite Cluster Strongest Class Correlation Cannabinoids Monoterpenes Sesquiterpenes 1 Class 2 CMPD1, CBDVA, CMPD5, CMPD6, CBDA, THCV, CMPD11, CMPD15, CMPD16, CMPD18 β-myrcene  2 Class 1, 3 CMPD14, CMPD17 p-cymene, γ-terpinene, p-cymenene, α-fenchene  3 Class 1, 3 CMPD2, CMPD7, CBGA, CMPD10, CMPD12 α-thujene, sabinene,  2-carene,  α-phellandrene,  3-carene, α-terpinene, β-phellandrene,  cis-β-ocimene, terpinolene, terpinen-4-ol  4   trans-β-ocimene, sabinene hydrate  5 Class 5 CMPD8, CBN, CMPD20 santolina triene,  α-pinene, camphene, camphene hydrate α-gurjunene,  sesquiterp-1, δ-cadinene 6 Class 8 CMPD4, CMPD13, CMPD19  δ-santalene, alloaromadendrene,  cis-β-farnesene, caryophyllene oxide 7 Class 8, 9 CBG, THCA,  Δ8-THC β-pinene Ylangene, γ-elemene, γ-muurolene,  γ-gurjunene, germacrene A, β-maaliene,  cis-α-bisabolene, germacrene B, δ-selinene 8 Class 4, 8, 9  D-limonene,  exo-fenchol,  trans-2-pinanol α-cubenene,  β-caryophyllene,  4,11-selinadiene,  β-selinene, α-selinene, valencene, gurjunene derivative,  guaia-3,9-diene,  selina-3,7(11)-diene  129 Metabolite Cluster Strongest Class Correlation Cannabinoids Monoterpenes Sesquiterpenes 9 Class 4 CMPD3, CMPD21 fenchone, linalool, endo-borneol,  α-terpineol β-elemene, α-guaiene, humulene, α-farnesene,  δ-amorphene,  δ-bulnesene,  β-sesquiphellandrene, guaiol, 10-epi-γ-eudesmol   Figure 5.6 Decision tree built from the HCA clusters using different cannabinoids and terpenes to identify the appropriate cluster. A potential tool for chemotaxonomic classification of Cannabis based on cannabinoids and terpenes.  5.3.4 Cannabinoids and terpenes: same strain, different producer comparison Canadian regulations under the original medical Cannabis regulations required producers to label their Cannabis products with brand names (Health Canada, 2016). Producers could choose to use their commercially traded strain name or with a company brand name that does not necessarily link to its strain name. After some investigation into 130 linking the brand names to their common strain names, two samples from different producers were identified as the same strain; nordle (CAN31 and CAN39). Both CAN31 and CAN39 were identified in the high THC:high CBD class (orange group) based on their cannabinoid profiles. The cannabinoid and terpene profiles are summarized in Table 5.3. Based on the ratio of total THC:CBD in the strains of 0.86 in CAN31 and 0.73 in CAN39, the expression of THC to CBD is consistent. Minor cannabinoid variance describes the variance of all cannabinoids other than THCA, THC, CBDA and CBD, which was determined to be 0.03. The three major monoterpenes were also consistent between the two materials, where the relative proportions of them vary less than 10%. Variance of terpenes not expressed in all cannabinoid clusters summarizes the variance of the terpenes described in groups 2 to 5 in Chapter 4, which was 0.05 between CAN31 and CAN39. Table 5.3 Summary of the between strain variance of two strains purchased from different producers.  CAN31 CAN39 Cannabinoids THC (% w/w) 6.72 5.75 CBD(% w/w) 7.84 7.87 Ratio 0.86 0.73 Total Variance 0.77 Minor Cannabinoid Variance 0.03 Terpenes Major β-myrcene 47.6 % 55.5 % D-limonene 29.6 % 21.7 % α-pinene 9.2 % 5.1 % Total Variance 73.1  Variance of Terpenes not expressed in all clusters 0.05  Figure 5.7 illustrates the expression of the cannabinoids between the two samples of nordle. Aside from the variance in THC and CBD levels, the expression of different cannabinoids is consistent. In sample CAN31, there is a higher prevalence of THC and CBD compared to CAN39 indicating some loss of acidic cannabinoids due to processing, handling or storage.  131  Figure 5.7 Bar graph comparing the content of each cannabinoid detected the nordle samples provided by two different producers.  Figure 5.8 illustrates the terpene profiles of the two samples of nordle. The profiles are relatively consistent. There are slight variations in the proportions of the major terpenes, but there are no obvious changes in expression. Both strains are β-myrcene-dominant with high proportions of limonene. 012345678910UK1UK2CBDVAUK3UK4UK5UK6CBDAUK7UK8CBGAUK9UK10THCVUK11CBDUK12CBGUK13UK14UK15UK16UK17THCACBNUK18THC8-THCCBCUK19UK20UK21Concentration (%w/w)can31can39132  Figure 5.8 Terpene headspace profiles comparing the nordle samples from two different producers.   0102030405060santolina trienea-thujenea-pinenecampheneSabineneB-pineneB-myrcene2-carenea-phellandrene3-carenea-terpinenep-cymeneD-limoneneB-phellandrenetrans-B-ocimenecis-B-Ocimenegamma-terpineneZ-sabinene hydrateterpinolenefenchonep-Cymenenelinaloolexo-fencholtrans-2-pinanola-Fenchenecamphene hydrateendo-borneolterpinen-4-ola-terpineola-cubeneneYlangenecopaeneB-cubebeneB-elemenecaryophyllenea-Santalenegamma-Elemenea-guaienea-gurjungeneunknown1humulenealloaromadendrenecis-B-farnesene(Z,Z)-a-Farnesenegamma-muurolenea-Amorphene4,11-selinadieneB-selinenegamma-gurjunenea-selinenea-BulneseneGermacrene AValencened-cadinenea-Gurjunene derivativeB-SesquiphellandreneB-maalieneguaia-3,9-dieneselina-3,7(11)-dienecis-a-bisaboleneGermacrene Bd-SelineneCaryophyllene oxideguaiol10-epi-gamma-Eudesmola-EudesmolBulnesolPeak Area (%)can31can39133 5.3.5 Cannabinoids and terpenes: lot-to-lot comparisons A total of nine new lots were acquired for comparison of between lot expression of cannabinoids and terpenes. A summary describing the cannabinoids in terms of total THC, total CBD, ratio of the two major cannabinoids and total variance are in Table 5.4. Eight of the nine strains were high THC strains with limited expression of CBD, while Nebula CBD (CAN19/CAN63) was a THC/CBD hybrid strain. There was a reduction in total cannabinoids detected in several strains, while ice cream, acapulco gold and bubba kush were very consistent. The ratio of THC/CBD in the lots of Nebula CBD were consistent. Table 5.4 Variance of cannabinoids observed between two lots of the same strain from the same producer. Strains THC (% w/w) CBD (% w/w) THC:CBD Ratio Total Variance Minor Cannabinoid Variance Ice Cream CAN15 17.3 <DL* >100 0.72 0.10 CAN62 18.3 <DL >100 Sensi Star CAN17 19.1 0.15 >100 12.6 0.87 CAN64 16.4 0.01 >100 Wappa CAN18 17.5 0.11 >100 4.8 0.05 CAN61 14.6 <DL >100 Nebula CBD CAN19 5.0 9.7 0.52 14.5 0.07 CAN63 2.7 5.8 0.47 Spoetnik CAN21 16.0 0.03 >100 11.1 0.27 CAN60 11.6 <DL >100 Chronic Thunder CAN22 12.1 0.04 >100 11.2 0.01 CAN56 9.4 <DL >100 Acapulco Gold CAN24 15.5 0.03 >100 2.3 2.01 CAN57 16.3 <DL >100 Blueberry Lambsbread CAN25 13.7 0.06 >100 14.0 0.01 CAN58 8.6 <DL >100 Bubba Kush CAN26 18.6 0.03 >100 0.12 0.04 CAN55 18.9 <DL >100 *<DL denoted below detection limit (DL) The total variance ranged from 0.12 to 14.5, indicating considerable differences in variance within the data set. In many of these strains, this variance was observed due to the major cannabinoid content in all strains with the exception of acapulco gold (CAN24/CAN57), where the variance of THC/CBD was about 0.3, while the other cannabinoids was 2.01. Bar graphs comparing the profiles of all nine strains are shown in Figure 5.9. Acapulco gold had 134 a higher expression of CMPD2 and CBGA, with detectable amounts of CMPD10 and CMPD12, which were not detected in the first lot of this strain. 0510152025CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(a)0510152025CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(b)0510152025CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(c)135   024681012CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(d)02468101214161820CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(e)0246810121416CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(f)02468101214161820CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(g)136   Figure 5.9 Cannabinoid quantitative profiles comparing the lot-to-lot variance of nine Cannabis strains obtained two years apart. (a) SensiStar, (b) Ice cream, (c) Wappa, (d) Nebula CBD, (e) Spoetnik, (f) Chronic Thunder, (g) Acapulco Gold, (h) Blueberry Lambsbread, (i) Bubba Kush. Original strains are colored blue, while newly acquired are red.  Terpene profiles between lots are summarized in Table 5.5. The major terpenes expressed in each of the strains are consistent between the two lots, while some variance in proportions was observed, with the exception of acapulco gold. This strain originally produced a high proportion of limonene and the second lot contained high proportion of terpinolene, with expression of the other terpinolene-correlated terpenes discussed in Section 4.3.2. The variances ranged from 6.9 to 1618.9, where the variance between the lots of acapulco gold was more than eight times higher in comparison to the next highest strains. This was due to the loss of expression of limonene and the increased expression of terpinolene and its correlated group of terpenes. The Q-test was used to determine if the variance of acapulco 0246810121416CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(h)0510152025CMPD1CMPD2CBDVACMPD3CMPD4CMPD5CMPD6CBDACMPD7CMPD8CBGACMPD9CMPD10THCVCMPD11CBDCMPD12CBGCMPD13CMPD14CMPD15CMPD16CMPD17THCACBNCMPD18THC8-THCCBCCMPD19CMPD20CMPD21Concentration (%w/w)(i)137 gold was an outlier. With a QEXP of 0.881 and a QCRIT of 0.568 with a 95% confidence level, this sample was considered an outlier. In this case, the average variance observed between lots of 8 of the 9 strains was 58.7 with a standard deviation of 60.0. Table 5.5 Variance of terpene profiles between two lots of the same strain from the same producer. Strains Major Terpenes Total Variance Variance of terpenes not expressed in all classes Ice Cream  Limonene β-myrcene β-pinene 198.1 2.5 CAN15 52.2 15.9 8.0 CAN62 40.1 30.9 8.9 Sensi Star  Terpinolene β-myrcene α-pinene 39.2 35.9 CAN17 35.4 24.4 6.6 CAN64 27.9 25.3 8.3 Wappa  β-myrcene α-pinene β-pinene 6.9 1.0 CAN18 54.9 20.7 7.3 CAN61 51.8 22.0 6.8 Nebula CBD  Terpinolene β-myrcene α-pinene 33.2 30.4 CAN19 29.6 24.0 9.8 CAN63 23.0 25.1 11.5 Spoetnik  β-myrcene Terpinolene α-pinene 53.1 5.3 CAN21 40.1 16.3 10.2 CAN60 31.4 17.9 13.4 Chronic Thunder  Limonene β-myrcene β-pinene 22.2 0.7 CAN22 32.1 26.5 10.8 CAN56 33.5 32.0 8.3 Acapulco Gold  Limonene β-pinene α-pinene 1618.9 519.3 CAN24 48.9 14.7 12.4  Terpinolene β-pinene α-pinene CAN57 31.2 13.3 12.2 Blueberry Lambsbread  β-myrcene α-pinene Limonene 77.2 0.3 CAN25 49.3 21.6 12.9 CAN58 38.0 25.3 12.6 Bubba Kush  Limonene β-myrcene β-pinene 39.4 0.1 CAN26 39.4 20.5 11.1 CAN55 38.2 12.7 13.5   Wappa had the lowest variance in terpene expression, while acapulco gold had the highest variance. The terpene profiles of these strains are plotted as bar graphs for each lot to highlight the differences between the variances observed (Figure 5.10 and 5.11). In Figure 5.10, the expression of the major monoterpenes are consistent. The strain is dominant in β-myrcene, with the secondary monoterpene being α-pinene. This strain had between 5 and 10 138 % of its profile compose of β-pinene, limonene and cis-β-ocimene. The original lot of acapulco gold, described in Figure 5.11, was a limonene-dominant strain, while in the second lot of the strain was terpinolene-dominant. It is interesting to note that the other monoterpenes including α-pinene, β-pinene and β-myrcene were consistent between the lots. The new lot has considerable amounts of the terpinolene correlated terpenes and cis-β-ocimene.   139  Figure 5.10 GC headspace data of the monoterpene and sesquiterpene profiles of two lots of wappa which had the lowest variance of the nine strains evaluated. The original lot is colored blue, the newly acquired lot is red. 0102030405060santolina trienethujenea-pinenecampheneSabineneB-pineneB-myrcene2-carenea-phellandrene3-carenea-terpinenem-cymenecymeneLimoneneB-phellandrenetrans-B-ocimenecis-B-ocimenegamma-terpinenesabinene hydrateterpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanola-FencheneCamphene hydrateendo-borneolterpinen-4-ola-terpineola-cubenenea-ylangenecopaeneB-cubebeneB-elemeneB-caryophyllenea-Santalenegamma-Elemenea-guaienea-gurjungenesesquiterp-1humulenealloaromadendrenecis-b-farnesene(Z,Z)-α-Farnesenegamma-muurolenea-amorphene4,11-selinadieneB-selinenegamma-gurjunenea-selinenea-bulneseneGermacrene AValencened-cadinenea-gurjunene derivativeB-sesquiphellandreneB-maalieneguaia-3.9-dieneselina-3,7(11)-dienea-bisaboleneGermacrene Bdelta-selinenecaryophyllene oxideguaiol8-epi-gamma-eudesmolB-eudesmolbulnesola-bisabololPeak Area (%)140  Figure 5.11 GC headspace data of the monoterpene and sesquiterpene profiles of two lots of Acapulco gold which had the highest variance of the nine strains evaluated. The original lot is colored blue, the newly acquired lot is red.  0102030405060santolina trienethujenea-pinenecampheneSabineneB-pineneB-myrcene2-carenea-phellandrene3-carenea-terpinenem-cymenecymeneLimoneneB-phellandrenetrans-B-ocimenecis-B-ocimenegamma-terpinenesabinene hydrateterpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanola-FencheneCamphene hydrateendo-borneolterpinen-4-ola-terpineola-cubenenea-ylangenecopaeneB-cubebeneB-elemeneB-caryophyllenea-Santalenegamma-Elemenea-guaienea-gurjungenesesquiterp-1humulenealloaromadendrenecis-b-farnesene(Z,Z)-α-Farnesenegamma-muurolenea-amorphene4,11-selinadieneB-selinenegamma-gurjunenea-selinenea-bulneseneGermacrene AValencened-cadinenea-gurjunene derivativeB-sesquiphellandreneB-maalieneguaia-3.9-dieneselina-3,7(11)-dienea-bisaboleneGermacrene Bdelta-selinenecaryophyllene oxideguaiol8-epi-gamma-eudesmolB-eudesmolbulnesola-bisabololPeak  Area (%)141 5.4 Discussion While the majority of the metabolites between the different compound classes were not strongly correlated, Pearson correlations did highlight a few. Can23 was the only strain where CMPD20, sesquiterp-1 and δ-cadinene were detected, therefore very strong correlation coefficients were identified in this high THC-dominant strain. As well, CMPD3 was detected only in can14 and was strongly correlated with linalool (0.80) which was the most dominant strain of this monoterpene. Pearson correlation coefficients assume that data are normally distributed and there is a linear relationship between the metabolites, which can be impacted by extreme values within a dataset (Mukaka, 2012). While Pearson correlation coefficients are a good measure for initially evaluating the relationships between metabolites, additional algorithms were implemented to further evaluate the data structure and confirm the preliminary findings. Multi-block data fusion models, such as multiple factor analysis (MFA), are used in cases where multiple platforms are involved including instruments, concentration ranges, units or types of data structures (Abdi et al., 2013). They provide information on similarities between the different platforms, the relevance of each block and the underlying relationships between different variables (Boccard & Rudaz, 2014). Since cannabinoids and terpenes were collected on different instruments, and monoterpenes and sesquiterpenes volatilize at different capacities at 80 °C using headspace-GC analysis, these three metabolite classes were separated into individual data blocks to understand the underlying data structure. The data fusion models transform individual data blocks within a dataset prior to evaluating the global relationships between methods or metabolites (Boccard & Rudazz, 2014). The benefits of multiblock data fusion methods is that the values from each cluster will not dominate the model as each block is normalized by the equivalent of the standard deviation of that matrix, similar to autoscaling individual metabolites in principal component analysis (Abdi et al., 2013; 142 Boccard & Rudazz, 2014). The data fusion model further confirmed the relationship between the terpinolene correlated monoterpenes and a few low abundance cannabinoids: CMPD12, CMPD10 and CMPD7. Classification of strains by THC/CBD content does not sufficiently describe the phytochemical variance in the strains, therefore an unsupervised clustering approach was undertaken to provide unbiased clustering based on metabolite compositions. Agglomerative hierarchical clustering analysis (HCA) using Ward’s method was undertaken to cluster the strains using the fused data, which ensures the within group sum of squares is minimized (Crawford et al., 2015). Clustering analysis lacks the ability to identify the impact of the data structure on the clustering, therefore post hoc algorithms are necessary to understand the underlying metabolites impacting the cluster analysis (Meinicke et al., 2008). The SAM statistic and 1DSOM were both used to identify key metabolites that were used to develop a decision tree to re-classify the strains according to their phytochemical composition rather than their THC/CBD contents. This type of tree could provide a better understanding of strain clustering, metabolite variance and pharmacological significance of different strain clusters (Lewis et al., 2018). Chemotaxonomy is the classification of plants based on their chemical composition as opposed to phenotypic expression (Hegnauer, 1986). In the case of Cannabis strains, there is a considerable reliance on the phytochemical composition of the strains, which supports the need for better chemotaxonomic classification (Russo, 2011; Russo, 2019). The preliminary decision tree was developed to highlight the clustering of the strains, indicating which metabolites impact clustering and to provide a starting point for chemotaxonomic classification based on targeted metabolomics of Cannabis strains. There were many terpenes and cannabinoids involved in the chemotaxonomic classification, where future investigations into the pharmacological variance of these strains could provide insight into metabolite significance and be used for activity based chemotyping of Cannabis strains. 143 Lot-to-lot variance of nine strains evaluated the consistency of the phytochemical expression and metabolite relationships observed. With the exception of the strain acapulco gold, eight of the nine strains had consistent expression of cannabinoids and terpenes. This is only a comparison between two lots, where the data suggests that the growing conditions do not impact metabolite expression for these strains. There was variation in the total content of cannabinoids, where alterations in growth cycle have been shown to cause considerable variance in total metabolites (Fischedick et al., 2010; Aizpurua-Olaizola et al., 2016). In order to fully assess lot-to-lot variance of strains, it would be necessary to undertake power calculations to determine minimum sampling for statistically valid variance determinations to evaluate within-plant, within-lot and between-lot variance (Broadhurst & Kell, 2006; Blaise et al., 2016). Previous investigations have shown that Cannabis grown in controlled environments can have variations of less than 15% variance between lots of the same strain (Fischedick et al., 2010; Hazekamp & Fischedick, 2012).  The one strain that appeared to have considerable variation in metabolite expression between lots, acapulgo gold, showed consistent major cannabinoid content between the two lots, but the minor cannabinoids varied considerably. There was a higher prevalence of CBGA, with CMPD7 and CMPD12 observed in the newly acquired lot. When evaluating the terpene profile of this strain, there was considerable variation: originally a high limonene strain was now dominant in terpinolene and the presence of cis-β-ocimene and the other terpinolene-correlated cannabinoids was evident. This indicates two things: a) terpene expression of this strain has changed between lots and b) the presence of terpinolene is consistently associated with the other monoterpenes and minor cannabinoids identified in the original data set. After discussion with the producer, it was determined that the other strains were produced from the same mother plant to ensure minimal genetic variation between lots, acapulco gold had been regrown from a newly “popped” seed. The grower selected this chemotype for propagation, 144 which was a terpinolene-dominant chemotype. According to the producer, the phenotypic expression of this strain was consistent between the different seeds. The data confirmed that the major cannabinoid expression is consistent, while the expression of the terpenes appears to have more genetic variation in expression between seeds, which has previously been observed with propagation of hemp seeds (Booth et al., 2017). There are two common modes of seed production in Cannabis facilities (Clarke & Merlin, 2016b). There are “feminized seeds” which are produced by application of silver thiosulfate to female flowers to produce male flowers which are cross pollinated in order to maintain the genetic structure (Ram & Sett, 1982; Clarke & Merlin, 2016b; Chandra et al., 2017). Seeds can also be produced by crossing male and female flowers on two different plants to produce seeds, which would have a higher genetic variance (Clarke & Merlin, 2016b). In order to maintain genetic expression, producers propagate using cuttings from mother plants or tissue culture (Wang et al., 2009; Small, 2015; Chandra et al., 2017). Geranyl pyrohosphate (GPP) is produced via the methylerythritol phosphate (MEP) (Tholl, 2006; Tholl 2015). A data fusion model was selected to observe any potential underlying correlations between different metabolites to highlight any unique biosynthetic pathways links (Abdi et al., 2013; Boccard & Rudaz, 2014). By treating cannabinoids, monoterpenes and sesquiterpenes as separate data blocks, a relationship was observed between CMPD10, CMPD12 and terpinolene correlated monoterpenes which clustered separately from the majority of the Cannabis strains and warranted further investigation. The monoterpenes that were observed to be highly correlated with terpinolene were detected in high prevalence in both very high THC and high THC/CBD-containing strains. The aromatic descriptors for many of these terpenes are associated with citrus, pine and woody aromas, with terpinolene being one of the predominant terpenes in these strains (Breitmaier, 2006). There is minimal dependence on the cannabinoid composition in relation to terpinolene 145 therefore the selection pressures are more likely to be based on aromatic preferences and pharmacological experience (Clarke & Merlin, 2016b; Small 2015). The clear break in the expression of these monoterpenes, suggests the presence of a potentially unique pathway or enzyme contributing to their biosynthesis. Recent findings have indicated that many monoterpenes are produced from the GPP isomer neryl diphosphate (NPP) in tomatoes and lavender (Schilmiller et al., 2009; Adal et al., 2017). These monoterpenes include: α-phellandrene, β-phellandrene, α-terpinene, limonene, α-pinene, 2-carene, α-thujene, α-terpinolene, and 3-carene (Schilmiller et al., 2009; Adal et al., 2017). NPP also produces cis-CBGA, which can be used as a precursor in THCA synthesis (Degenhardt et al 2016; Fellermeier & Zenk, 1998). With the significant overlap in terpenes correlated with terpinolene in Cannabis compared with those found in the literature, it was hypothesized that some strains have an upstream biosynthetic pathway producing both NPP and GPP, resulting in the biosynthesis of terpinolene from NPP and CMPD12 was cis-CBGA produced from the condensation of NPP and olivetolic acid.  The correlation of cis-β-ocimene was consistently present at considerable levels with terpinolene, while it was also present without the production of significant amounts of terpinolene in some strains. The TPS from CsTPS6FN produces >97% cis-β-ocimene with GPP as the precursor (Booth et al., 2017). Based on the hypothesized biosynthetic pathways, the lack of terpinolene production would be due to the absence of NPP biosynthesis in strains containing this TPS. A proposed pathway describing this hypothesis can be found in Figure 5.12. CMPD12 appears to have the strongest correlations with terpinolene. The structure of this cannabinoid could explain the biosynthetic pathway of terpinolene-correlated terpenes, which are only present in a few Cannabis strains.146  Figure 5.12 Proposed pathway based on linked cannabinoid and terpenes for the biosynthesis of terpinolene and other correlated terpenes. X’s denote biosynthetic breaks. Strains presented in the order presented in the HCA clustering described in Figure 5.5: (left to right) can21, 33, 32, 17, 34, 19, 16, 14, 23, 42, 38, 41, 30, 35, 31, 28, 40, 20, 25, 18, 13, 27, 11, 36, 37, 24, 39, 12, 29, 26, 15, 22, 10. 0246810121416can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%)CsTPS6FNGPP NPPcis-β-ocimene00.020.040.060.080.10.12can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Concentration (% w/w)051015202530354045can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 00.511.522.53can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 00.20.40.60.811.21.41.61.8can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 00.511.522.533.544.5can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 01234567can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 00.10.20.30.40.50.60.70.80.9can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%) 00.511.522.5can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%)00.020.040.060.080.10.12can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%)00.20.40.60.811.21.41.61.8can21can33can32can17can34can19can16can14can23can42can38can41can30can35can31can28can40can20can25can18can13can27can11can36can37can24can39can12can29can26can15can22can10Peak Area (%)terpinolene 3-carene α-thujeneα-phellandrene β-phellandrene p-cymeneα-terpinene terpinen-4-ol -terpineneCMPD12HCA Strain Groups (Color Coded)HCA strain Groups (Color Coded)HCA Strain Groups (Color Coded) HCA Strain Groups (Color Coded) HCA Str in Groups (Color Coded)HCA Strain Groups (Color Coded) HCA Strain Groups (Color Coded) HCA Strain Groups (Color Coded)HCA Strain Groups (Color Coded) HCA Str in Groups (Color Coded) HCA Strain Groups (Color Coded)147 The sensitive analytical methods used for the targeted-untargeted metabolomic approach highlights the significance of evaluating low abundance metabolites (Sumner et al., 2003; Inui et al., 2016). Metabolomics has been highlighted as a hypothesis-generating tool to further investigate the relationship between CMPD12 and terpinolene. The consistency observed between strains with these metabolites provides putative evidence that they are associated with a specific biosynthetic pathway in Cannabis that warrants further investigation. The accumulation of terpinolene in Cannabis strains does not appear to be regulated in the same capacity in comparison with other terpenes. While most of the strains contained a proportion of the major monoterpenes: α-pinene, β-pinene, β-myrcene and limonene, the expression of terpinolene has a distinct break in biosynthesis between either high abundance or negligible expression. Cannabis phytochemistry presents a complex mixture of metabolites, that impacts aroma, perception and medicinal uses (Russo, 2011; Russo, 2019; Lewis et al., 2018; Gilbert & DiVerdi, 2018). In order to address the link between phytochemical diversity, genetic diversity and pharmacological significance of Cannabis strains, methodologies to cluster or classify the strains are necessary to improve patient access and potentially identify which strains can be used or substituted for medicinal uses based on their phytochemical composition (Russo, 2011; Lewis et al, 2018; Russo, 2019). Metabolomics presents an unbiased approach to classify strains into groups based on their similarities (Worley & Powers, 2013; Turi et al., 2015). Within this dataset, nine classes were identified based on varying cannabinoid and terpene compositions and the resulting decision tree was used to identify which class each strain belonged to. This allows for improved chemical descriptions of the stains compared to typical regulatory requirements describing only the major cannabinoids. Development of classification schemes based on phytochemical composition of strains will 148 ultimately provide consumers and patients with better access to reliable products in this growing and diverse marketplace. 149 Chapter 6: Untargeted Metabolomics for Discovery of Phytochemical Relationships: Exploration into the Terpinolene Pathway 6.1 Synopsis While chemometric models provide insight into the underlying relationships in metabolomic data sets, untargeted approaches lack sufficient data to fully elucidate metabolites that have been highlighted in the models (Turi et al., 2015; Hall, 2011; Dunn et al., 2012). In-depth investigations to characterize the metabolites of interest have been highlighted as one of the largest bottlenecks in metabolomic studies (Dunn et al., 2012). Several options for identification are possible through accurate mass determination, logical algorithms, MS/MS fragmentation patterns compared to databases or traditional isolation techniques (Dunn et al., 2012; Turi et al., 2015). While many of these options will provide putative identification of the metabolite, isolation is the most effective technique for definitively identifying the metabolites of interest. Metabolite identification can lead to the discovery of novel compounds, pathways and/or pharmaceutical products (Turi et al., 2015; Tugizimana et al., 2013; Dunn et al, 2012). The link between several low concentration cannabinoids and terpinolene-correlated terpenes observed in Chapter 5 suggests a novel biosynthetic pathway that has yet to be explored in Cannabis.  I hypothesized that the biosynthetic pathway for the production of high terpinolene strains is dependent on the presence of the monoterpene precursor NPP, an isomer of GPP. Isolating and elucidating the more strongly correlated cannabinoid, CMPD12, will provide valuable information to elucidate the potential biosynthetic pathway required for terpinolene production in strains and identify its relation to domestication syndrome and clandestine breeding of Cannabis strains. 150 6.2 Experimental 6.2.1 Test samples The six strains with high terpinolene profiles were selected: can16, can17, can19, can21, can32 and can33. All strains were evaluated for terpene quantitation. Previous cannabinoid data indicated that can21 had one of the highest contents of CMPD12, therefore was selected for the isolation experiment. Intermediate isolates were stored in the refrigerator at 4 °C to ensure compound stability. 6.2.2 Reagents and calibration standards HPLC grade methanol was purchased from VWR International (Mississauga, ON, Canada). Cannabis Terpene Mix A and Mix B containing 20 and 15 terpenes, respectively, at 2000 µg/mL in methanol were purchased from Sigma Aldrich (Oakville, ON, Canada). Cannabis Terpene Mix A contained: α-pinene, camphene, β-pinene, 3-carene, α-terpinene, limonene, γ-terpinene, fenchone, fenchol, camphor, isoborneol, menthol, citronellol, pulegone, geranyl acetate, α-cedrene, α-humulene, nerolidol, cedrol and α-bisabolol. Cannabis Terpene Mix B contained: β-pinene, 3-carene, p-cymene, limonene, terpinolene, linalool, camphor, borneol, α-terpineol, geraniol, β-caryophyllene, cis-nerolidol, β-eudesmol and phytol. All standards were stored according to the manufacturer’s recommendation until use. Dilutions of the standard mixtures were prepared for the calibration standards with concentrations of 1, 2.5, 10, 25 and 100 µg/mL of each standard. Standard mix A and standard mix B calibration curves were prepared separately as there were several identical terpenes in both solutions. 6.2.3 Terpinolene quantitation Cannabis flowers were ground with liquid nitrogen with a mortar and pestle immediately prior to extraction. All test materials were prepared in triplicate. Aliquots of 100.0 151 mg of ground flower were placed into a 50 mL amber centrifuge tube and extracted with 5 mL of methanol. Samples were vortexed for 30 seconds and then subjected to 20 minutes in a sonicating bath. Following extraction, the samples were centrifuged at 4500 g for 5 minutes and filtered through a 0.2 µm PTFE filter into a 2 mL GC vial. Extracts were analyzed by GC-MS. An aliquot of 1 µL of the methanol extract was injected into the GC inlet at 230 °C with a split ratio of 1:20. Separation occurred on an Agilent J&W DB-5MS 20 m x 180 µm ID, 0.18 µm film thickness column with helium as the carrier gas at 1.3 mL/min. The temperature gradient within the column started at 50 °C for 3 minutes, then ramped to 170 °C at 5 °C/min. Electron impact ionized the terpenes and a mass range of m/z 50 to 500 was acquired. Peak areas for terpenes were measured and used for quantitation against calibration curves prepared with the mixture standards. Monoterpenes that were not present in the standard mixture were quantified against limonene and sesquiterpenes were quantified against β-caryophyllene. 6.2.3.1 Quantitation validation Method performance was assessed for three major terpenes: α-pinene, limonene and β-caryophyllene based on being located at different locations in the chromatographic separation of authentic Cannabis strains with varying terpene concentrations. Selectivity was determined by evaluating the retention time comparisons between samples and standards with an error of less than ± 0.2 minutes being acceptable. Metabolite selectivity was also verified by evaluating the mass spectra for peak purity based on relative matching scores using the NIST spectral database with scores greater than 900 being acceptable. The linearity for each calibration curve was assessed by evaluating the coefficient of determination (r2). Values greater than 0.99 were considered suitable. Repeatability was assessed by determining the % RSD of triplicate samples for the three major terpenes in authentic materials with varying terpene concentrations.  152 6.2.4 CMPD12 isolation Five grams of can21 dried flower were ground using liquid nitrogen with a mortar and pestle. The ground material was immediately extracted with 50 mL of methanol using sonication for 30 minutes. The sample was centrifuged at 4500 g for 5 minutes and the supernatant placed into a round bottle flask. The residue was extracted three additional times with 50 mL of methanol and the supernatants were pooled into the round bottom flask. The extract was dried under vacuum in reduced light at 35 °C yielding 2.07 grams of resinous extract. The extract was fractionated on a Grace Reveleris flash chromatography system using a 12 gram reversed phase C18 column (Buchi, New Castle, DE). Isocratic elution composed of 30:70 (% v/v) 10 mM ammonium formate pH 3.6:acetonitrile was used to fractionate the extract. A dynamic fraction collector collected fractions based on peaks in the UV chromatogram at 220 nm. The fractions containing CMPD12 were collected into a round bottom flask and the organic mobile phase was removed under vacuum at 35 °C. The remaining water was removed by lyophilization. The residue was dissolved in methanol and further isolated with an Agilent 1100 semi-preparative HPLC equipped with an Agilent Zorbax SB-C18 column (9.4 x 50 mm, 5 µm). The cannabinoids were separated with gradient elution with a mobile phase composed of 10 mM ammonium formate pH 3.6 and acetonitrile. With a flow rate of 8 mL/min, the gradient was as follows: 0 to 4 min: 52% B, 4 to 11 min: 52-55 %B, 11-12 min: 95 %B. Fractions containing CMPD12 were collected into a round bottom flask and evaporated to dryness at 35 °C. The final fraction was separated by adjusting the pH of the mobile phase by using 0.1% formic acid in place of the buffer with the same chromatographic conditions. CMPD12 isolated fractions were pooled into a round bottom flask, organic solvent was removed under vacuum and the final aqueous solution was freeze 153 dried. The remaining isolate (2.6 mg) was dissolved in CDCl3 and analyzed by HPLC-UV-MS and NMR for structure elucidation. 6.2.5 Structure elucidation  6.2.5.1 NMR spectroscopy NMR spectra were obtained at 25 °C with TSP as an internal standard at 400.20 and/or 100.64 MHz using a Bruker Avance III HDTM 400 MHz Spectrometer (Bruker, Billerica, MA). Proton (1H) and carbon (13C) one-dimensional experiments were collected in addition to the following 2-dimensional experiments: COSY, HSQC and HMBC. All spectra were acquired using standard acquisition parameters. 6.2.5.2 LC-MS acquisition LC-MS and MS/MS analyses were performed on an Agilent 1290 LC system coupled to a 6420 mass spectrometer (Agilent Technologies). Chromatographic separation of the cannabinoids was consistent with the previous validated method described in Chapter 2. Optimization of the MS parameters were undertaken in-house using THCA as the reference standard. Mass spectrometric detection was undertaken in positive polarity with a capillary voltage of 4000 V, gas temperature of 350 °C, gas flow of 10 L/min, and nebulizer pressure of 40 psi. The full scan acquisition was from 100 to 500 m/z with a fragmentor voltage of 80 V and a cell accelerator voltage of 4 V. Product ion scans were acquired with preselected precursor ions based on the full scan acquisitions. The collision energy used was 20 V. 6.3 Results 6.3.1 Terpene quantitation Quantitative measurements of terpenes present in the samples were acquired to verify the results of the headspace profiling. Terpenes were quantified against the terpene standard 154 mixes purchased from Sigma Aldrich. Terpenes without reference standards available were quantified against limonene and β-caryophyllene for monoterpenes and sesquiterpenes, respectively. Terpenes were quantified in all 33 samples in the Cannabis data set described in Chapters 3 to 5. The total monoterpene and total sesquiterpenes measured in each sample are summarized in Table 6.1. Monoterpenes ranged from 0.60 to 3.14 mg/g with an average content of 1.56 mg/g. Sesquiterpenes ranged from 1.46 to 11.35 mg/g with an average content of 4.76 mg/g. Pearson correlation coefficients were calculated between the concentration of monoterpenes, sesquiterpenes and total cannabinoid content as shown in Figure 6.1. There was a reduction in sensitivity in comparison to the headspace analysis due to the solvent extraction. Several of the low concentration terpenes, such as β-phellandrene, were not detected in samples that were previously detected with headspace analysis.   155 Table 6.1 Total monoterpene and sesquiterpene concentrations measured in all thirty-three Cannabis strains. Error ± SEM (n=3) Strain Total Monoterpenes Concentration (mg/g) Total Sesquiterpenes Concentration (mg/g) CAN10 1.07 ± 0.05 4.59 ± 0.27 CAN11 1.19 ± 0.03 5.01 ± 0.12 CAN12 1.33 ± 0.02 5.15 ± 0.08 CAN13 1.24 ± 0.03 3.82 ± 0.04 CAN14 0.99 ± 0.02 4.05 ± 0.07 CAN15 1.31 ± 0.03 6.81 ± 0.17 CAN16 2.61 ± 0.03 3.67 ± 0.07 CAN17 3.14 ± 0.06 4.46 ± 0.04 CAN18 1.48 ± 0.04 6.48 ± 0.08 CAN19 2.34 ± 0.07 3.12 ± 0.08 CAN20 0.98 ± 0.02 3.97 ± 0.09 CAN21 2.00 ± 0.05 4.31 ± 0.17 CAN22 1.47 ± 0.03 6.59 ± 0.17 CAN23 1.40 ± 0.04 3.43 ± 0.10 CAN24 2.35 ± 0.04 11.35 ± 0.14 CAN25 1.60 ± 0.05 4.65 ± 0.07 CAN26 2.03 ± 0.06 9.85 ± 0.17 CAN27 2.84 ± 0.07 5.33 ± 0.10 CAN28 0.60 ± 0.02 1.46 ± 0.17 CAN29 2.01 ± 0.05 8.11 ± 0.05 CAN30 1.05 ± 0.02 2.95 ± 0.17 CAN31 1.27 ± 0.03 5.84 ± 0.24 CAN32 2.11 ± 0.05 4.71 ± 0.04 CAN33 2.06 ± 0.05 2.49 ± 0.05 CAN34 1.47 ± 0.03 3.04 ± 0.05 CAN35 0.69 ± 0.01 2.94 ± 0.05 CAN36 2.10 ± 0.08 5.45 ± 0.22 CAN37 1.07 ± 0.02 5.04 ± 0.05 CAN38 1.07 ± 0.04 2.35 ± 0.02 CAN39 0.93 ± 0.02 5.51 ± 0.10 CAN40 1.32 ± 0.03 4.50 ± 0.06 CAN41 1.01 ± 0.01 3.33 ± 0.04 CAN42 1.22 ± 0.03 2.84 ± 0.03  156  Figure 6.1 Correlation plots evaluating the relationship between total cannabinoids content and terpenes. (a) Correlation plot of total cannabinoid content versus total monoterpene content and (b) correlation plot of total cannabinoid content versus total sesquiterpene content.  The average content of monoterpenes present in terpinolene strains was 2.38 mg/g, of which five of six are in the top 25% of the data. The content of the four major monoterpenes; α-pinene, β-myrcene, limonene and terpinolene, are shown in Figure 6.2. The distribution of these terpenes closely resembles that shown previously with headspace analysis. The first three monoterpenes: α-pinene, β-myrcene, limonene are present across each cannabinoid class with minimal variation across the dataset. The terpinolene expression appears to remain present in quantifiable amounts in the orange and blue (high THC/CBD and very high THC) 0510152025300.00 1.00 2.00 3.00 4.00Total Cannabinoid Concentration (%w/w)Total Monoterpene Concentration (mg/g)(a)ρ=0.590510152025300.00 5.00 10.00 15.00Total Cannabinoid Concentration (%w/w)Total Sesquiterpene Concentration (mg/g)(b)ρ=0.47157 cannabinoid classes, with the six strains previously identified as high terpinolene containing the highest content of this monoterpene. 158    Figure 6.2 Monoterpene contents of (a) α-pinene, (b) β-myrcene, (c) limonene and (d) terpinolene across the Cannabis dataset. Error bars represent SEM (n=3). Strains a colored according to their cannabinoid class described in Chapter 3.  00.20.40.60.811.21.4CAN34CAN38CAN30CAN35CAN19CAN16CAN14CAN39CAN41CAN31CAN28CAN40CAN23CAN33CAN12CAN22CAN10CAN20CAN37CAN42CAN11CAN25CAN13CAN27CAN36CAN24CAN21CAN29CAN15CAN18CAN32CAN17CAN26Concentration (mg/g)(a)00.050.10.150.20.250.30.350.40.45CAN34CAN38CAN30CAN35CAN19CAN16CAN14CAN39CAN41CAN31CAN28CAN40CAN23CAN33CAN12CAN22CAN10CAN20CAN37CAN42CAN11CAN25CAN13CAN27CAN36CAN24CAN21CAN29CAN15CAN18CAN32CAN17CAN26Concentration (mg/g)(b)00.10.20.30.40.50.60.70.8CAN34CAN38CAN30CAN35CAN19CAN16CAN14CAN39CAN41CAN31CAN28CAN40CAN23CAN33CAN12CAN22CAN10CAN20CAN37CAN42CAN11CAN25CAN13CAN27CAN36CAN24CAN21CAN29CAN15CAN18CAN32CAN17CAN26Concentration (mg/g)(c)00.20.40.60.811.2CAN34CAN38CAN30CAN35CAN19CAN16CAN14CAN39CAN41CAN31CAN28CAN40CAN23CAN33CAN12CAN22CAN10CAN20CAN37CAN42CAN11CAN25CAN13CAN27CAN36CAN24CAN21CAN29CAN15CAN18CAN32CAN17CAN26Concentration (mg/g)(d)159 Monoterpenes quantified in the terpinolene-dominant strains were plotted in Figure 6.3a-f. Terpinolene is in the highest abundance in all six strains. Three other non-terpinolene strains were plotted as a comparison can39, can22 and can18 in Figure 6.3g-i. These strains were all classified in the orange and blue classes in close proximity to the different high terpinolene strains but did not have quantifiable levels of terpinolene. The monoterpenes which were previously identified as correlated with terpinolene are consistent in the quantitative data. The three representative strains below do not contain significant amounts of these monoterpenes, which is consistent across the dataset. This data confirms the previous findings with headspace analysis is consistent with that found in the quantitative monoterpene data.   00.20.40.60.811.2α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(a)00.20.40.60.811.2α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(b)00.050.10.150.20.250.30.350.40.45α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(c)00.050.10.150.20.250.3α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(d)160   Figure 6.3 Monoterpene profiles of the six terpinolene-dominant strains (a) can16, (b) can17, (c) can19, (d) can21 (e) can32, (f) can 33 in comparison to three representative non-terpinolene-dominant strains (g) can39, (h) can22, (i) can18. Error bars represent SEM (n=3).   Similarly, the content of sesquiterpenes in the six high terpinolene strains as well as the three representative strains from those cannabinoid classes do not indicate any significant 00.050.10.150.20.250.30.350.4α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(e)00.050.10.150.20.250.3α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(f)00.020.040.060.080.10.120.140.16α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(g)00.050.10.150.20.250.3α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene…terpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(h)00.050.10.150.20.250.30.35α-thujeneα-pinenecamphenesabineneβ-pineneβ-myrceneα-phellandrene3-careneα-terpinenep-cymenelimonenecis-β-ocimeneγ-terpinenesabinene hydrateterpinolenefenchonep-cymenenelinaloolexo-fencholtrans-2-pinanolα-fenchenecamphorendo-borneolterpinen-4-olα-terpineolConcentration (mg/g)(i)161 correlations between the sesquiterpenes, terpinolene or cannabinoids. These profiles are shown in Figure 6.4. 00.10.20.30.40.50.6α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(a)00.10.20.30.40.50.6α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(b)00.050.10.150.20.250.30.35α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(c)162  00.10.20.30.40.50.6α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(d)00.10.20.30.40.50.60.7α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(e)00.050.10.150.20.250.3α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(f)163   Figure 6.4 Sesquiterpene profiles of the six terpinolene-dominant strains (a) can16, (b) can17, (c) can19, (d) can21 (e) can32, (f) can 33 in comparison to three representative non-terpinolene-dominant strains (g) can39, (h) can22, (i) can18. Error bars represent SEM (n=3). 00.10.20.30.40.50.6α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(g)00.10.20.30.40.50.60.70.8α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(h)00.20.40.60.811.21.4α-ylangeneα-cubebeneβ-cubebeneγ-elemeneβ-caryophylleneα-guaieneα-gurjungenehumulenecis-β-farneseneγ-muuroleneα-amorphene4,11-selinadieneβ-selineneγ-gurjuneneα-selineneα-bulnesenegermacrene Avalenceneα-gurjunene derivativeβ-sesquiphellandreneβ-maalieneguaia-3,9-dieneselina-3,7(11)-dieneα-bisabolenegermacrene Bδ-selinenecaryophyllene oxideguaiol8-epi-γ-eudesmolβ-eudesmolbulnesolα-bisabololConcentration (mg/g)(i)164 6.3.1.1 Terpene quantitation: validation All peaks in the reference standards matched the retention times of the peaks in the samples. The 95 % confidence interval of retention time in 5 samples prepared in triplicate was for α-pinene was 4.952 ± 0.003 minutes, for limonene was 7.6913 ±0.003 minutes and β-caryophyllene was 18.5659 ± 0.0008 minutes. The MS spectra were used to confirm selectivity by evaluating peak identity and purity with relative matches greater than 900 for α-pinene, limonene and β-caryophyllene for all reference standards and authentic plant materials. The coefficients of determination (r2) values for all calibration curves ranged from 0.998 to 0.999, confirming the linear response of the terpenes within the range evaluated. Relative standard deviations for the quantitative results for α-pinene with a concentration range of 0.05 to 1.22 mg/g ranged from 0.21 to 12.9 % RSD, with an average of 4.1 % RSD. The RSDs for limonene with concentrations ranging from 0.05 to 0.75 mg/g ranged from 0.40 to 12.9 % RSD with an average of 5.11 % RSD. The RSDs for β-caryophyllene with concentrations from 0.09 to 2.97 mg/g ranged from 0.57 to 15.4 % RSD with an average of 3.8 % RSD. Given the performance of most samples being less than 10%, the few samples with RSDs greater than 10% are likely due to sampling or inhomogeneity issues rather than method performance. 6.3.2 CMPD12 isolation and characterization The bulk extract obtained from the methanol extraction of can21 was highly viscous and oily, with a high content of THCA and other resinous constituents. The initial fractionation on the flash system was used to remove the cannabinoids from the other fat-soluble materials, for which a reversed phase preparatory column was best suited. A total of 82 fractions were collected from this analysis using the UV wavelength at 220 nm to automatically adjust for new peaks eluting from the system. Eight fractions were found to contain CMPD12 from this 165 separation and these were combined for analysis, but there were many additional cannabinoids present in the fraction including THCA.  Semi-preparative HPLC was then used to collect more purified fractions of CMPD12. The first separation with 10 mM ammonium formate at pH 3.6 mimicked chromatographic separation conditions. The UV chromatogram of this separation is shown in Figure 6.5. CMPD12 eluted between 7-10 minutes and the four fractions were combined. LC-MS identified several minor co-eluting cannabinoids with CMPD12 in this fraction, with the highest abundance contaminant being a neutral cannabinoid. By adjusting the pH of the mobile phase, it was possible to isolate CMPD12 from this co-eluting cannabinoid (Figure 6.6).  Figure 6.5 Semi-preparative HPLC separation of the cannabinoid fraction obtained after flash chromatography to isolate CMPD12 at 220 nm.  166  Figure 6.6 Final semi-preparative HPLC separation of CMPD12 from other minor contaminants at 220nm to isolate CMPD12.  The total amount of CMPD12 isolated was 2.6 mg. This was dissolved in CDCl3 and subjected to NMR spectroscopy and HPLC-UV-MS. NMR data summarized in Table 6.2 indicate that this cannabinoid closely resembles THCA (Choi et al., 2004). The terpene sidechain of this molecule is identical as summarized in the table. The presence of shifts at 6.26 (s) and 6.42 (m) confirm CMPD12 as an acidic cannabinoid as well as the UV maximum at 223 nm. Previous characterization of THCA shows that there are overlapping shifts of the 3” and 4” of the polyketide sidechain using NMR (Choi et al., 2004). Therefore, LC-MS was also employed. The cannabinoid had a protonated molecular ion at m/z 345, indicating a loss of CH2 from THCA which has a molecular ion of m/z 359. Tandem MS spectra of THCA and CMPD12 were compared (Figure 6.7). The base peak observed in both spectra was the loss of water [m/z -18]. The second most abundant fragment in THCA was the loss of the monoterpene sidechain [m/z -140]. CMPD12 also produced the loss of 140, indicating that the 167 loss of CH2 does not occur at the monoterpene sidechain. The earlier elution time in the chromatographic separation confirms a less hydrophobic structure (Figure 6.8), therefore confirming that the structure of this compound is THCA-C4 which has a butyl sidechain on the polyketide, previously characterized by Harvey (1976). The structure is described in Figure 6.9.   168 Table 6.2 Carbon (δC) and proton (δH) NMR spectroscopic data (400 MHz, CDCl3) for CMPD12, identified as THCA-C4. Position THCA-C4 (CMPD12) δC, type δH,(J in Hz) 1 33.9, CH 3.25, m 2 123.4, CH 6.42, bs 3 133.4, C  3-Me 23.3, CH3 1.70, s 4 31.4, CH2 2.20, m 5 25.5, CH2 1.92, m 6 46.5, CH 1.70, m 7 78.9, C  8 27.8, CH3 1.48, s 9 19.8, CH3 1.13, s 1’ 109.7, C  2’ 163.5, C  3’ 101.9, C  4’ 147.8, C  5’ 113.3, CH 6.26, s 6’ 159.9, C  1” 37.0, CH2 2.96, m & 2.80, m 2” 33.1, CH2 1.58, m 3” 29.8, CH2 1.36, m 4” 14.7, CH3 0.93, t (7.3) 2’-OH  12.19, s COOH 174.0   169  Figure 6.7 Product ion spectra from (a) THCA with a molecular ion of m/z 359 and (b) THCA-C4 (CMPD12) with a molecular ion of m/z 345. (a) (b) - m/z 18, H2O - m/z 98 - m/z 140 - m/z 18, H2O - m/z 140 - m/z 219 170  Figure 6.8 Chromatographic separation of CMPD12 (THCA-C4; blue) and THCA (red) with detection at 220 nm.   Figure 6.9 Structure of CMPD12, identified as THCA-C4 with a butyl sidechain on the polyketide moiety.  6.3.3 LC-MS characterization of correlated cannabinoid (CMPD7) The LC-MS profile of can21 was collected to putatively identify the other correlated cannabinoids: CMPD7 and CMPD10. The concentration and close elution with other cannabinoids of CMPD10 was insufficient for accurate determination with LC-MS. CMPD7 171 had a protonated molecular ion at m/z 331, indicating a loss of C2H4 from THCA, therefore was putatively identified as THCVA with a propyl polyketide sidechain. This cannabinoid is commercially-available from Cerilliant, therefore the standard was acquired to confirm the structure. The THCVA standard had a consistent elution time, molecular mass and fragmentation pattern to CMPD7 (Figure 6.10). Therefore, based on these data there is a strong correlation with higher concentrations of acidic THC cannabinoids with shorter polyketide sidechains with terpinolene and its related terpenes.  Figure 6.10 Product ion scan of THCVA with a molecular ion at m/z 331 used to confirm the identity of CMPD7.  6.4 Discussion Headspace analysis represents the volatilized metabolites of a sample at a specific temperature and can provide valuable information on aroma characteristics and metabolite profiles, but during this research it was found to be unsuitable for quantitative determination of terpenes as the headspace of the standards diluted in methanol produced non-linear responses. In this case, it was necessary to evaluate the findings of the headspace analysis by comparing with solvent extraction methods for quantitative confirmation of terpene profiles - m/z 18, H2O - m/z 140 - m/z 98 172 in Cannabis flowers. The solvent extraction method did produce similar results to those found in the headspace analysis. For example, can27 was α-pinene-dominant in both methods, can14 was linalool-dominant and the six terpinolene-dominant strains had consistent profiles with terpinolene and the remaining correlated monoterpenes. The limitation of the solvent extraction method was a lower sensitivity. As the injection volume was much lower, it was not possible to detect the low abundance underlying terpenes that could be detected and identified in the headspace method. Therefore, both methods produce different types of data but are both suitable to identify unique relationships between different metabolites and metabolite classes. In the case of headspace analysis, it was possible to detect unique, low abundance metabolites which were found in cannabinoid classes and represents the aromatic expression of strains, while the solvent extraction provided quantitative expression comparisons of different metabolites to understand the phytochemical profiles within the strains. Interestingly, the β-caryophyllene-to-humulene ratio detected in the headspace method was approximately 4.6 ± 0.8, while the quantitative results it was 2.2 ± 0.5, which is closer to that described by the expression of the terpene synthase for these two sesquiterpenes (Booth et al., 2017).  The quantitative results further substantiate the findings described in the previous chapters relating the correlations between terpinolene and the monoterpenes: α-thujene, α-phellandrene, 3-carene, α-terpinene, cymene, γ-terpinene, and cymenene. A few of the other monoterpenes were not detected with the solvent extraction method. The quantitative data provided additional information related to total monoterpene and sesquiterpene abundance within Cannabis flowers. It was postulated that there may be a relationship between the total cannabinoid content and total monoterpenes as these both rely on the presence of the same precursor (Degenhardt et al., 2017). The data identified a weak positive correlation between the two. It should be noted that the cannabinoid quantitation and headspace analysis was 173 performed within a month of acquisition of the samples, while terpene quantitation was undertaken after about 18 months of storage. Due to legal restrictions, the samples (whole flowers) were stored at room temperature in a locked safe. Minimal handling was undertaken to ensure the trichomes were intact, but there may have been some losses of the terpenes during storage (Ross & Elsohly, 1996; Small & Naraine, 2016). It would be expected that the volatilization of the terpenes would be similar across all strains, although this may be impacted by container type, moisture content and strain quality, it does provide some insight into comparisons between strains (Small & Naraine, 2016). Due to this limitation, it is not best to compare the metabolites themselves rather than the total amounts as these may not be accurate representations of the original samples. The data still substantiate the investigations into CMPD12 correlation with terpinolene, which was the original purpose of comparing the headspace and solvent extraction data. THCA-C4 and THCVA are biosynthesized from CBGA-type precursors which are products of GPP and a polyketide with varying lengths of the carbon sidechain on the polyketide (THCVA = propyl, THCA-C4 = butyl) (Shoyama et al., 1984; Smith, 1997). Terpinolene and these cannabinoids were only expressed in the very high THC and high CBD/THC hybrid strains, which are high total cannabinioid strains and five of the six strains had higher total monoterpene contents. The increased content of these additional THCA-like cannabinoids suggests that the production of precursors is overexpressed in these strains. Biosynthesis of secondary metabolites in plants requires energy (Hartmann, 2007), therefore the overabundance of precursors synthesized and accumulated in these strains triggers the switching on of the biosynthesis of terpinolene and its related monoterpenes. In the MEP pathway, increased expression of the enzyme DXR in peppermint resulted in 50% increased yield of essential oil (Mahmoud & Crouteau, 2001).  Therefore, the expression of this enzyme may be increased in terpinolene-dominant strains. The correlated cannabinoids have shorter 174 polyketide chains that may require less energy for the plant to produce in response to the stress conditions. This results in the production of terpinolene and these other cannabinoids that are not as prevalent in other Cannabis strains. Evaluation of cannabis strains focusing on chemotyping and classifying according to ‘indica’/’sativa’ designations has identified terpinolene to be a common terpene associated with ‘sativa’ type strains (Hazekamp & Fischedick, 2012; Hazekamp et al., 2016). Based on the six terpinolene-dominant strains evaluated in this dataset, two were designated as ‘indica’, one was designated as ‘sativa’ and three were designated as ‘hybrid’ strains. While the definition of these designations is vague and unspecific (McPartland, 2018), this confirms the complexity of strain designations and pharmacological associations of Cannabis phytochemistry. Hazekamp & Fischedick (2012) previously identified correlations between CBG, THCV and terpinolene, but only compared the phytochemical composition of two strains. By evaluating a larger dataset, it was possible to confirm these findings with additional THC-like cannabinoids (THCA-C4) and identify relationships with total cannabinoid content, breeding and domestication. The use of metabolomics as a hypothesis-generating tool has been highlighted in this study (Turi et al., 2015). The clustered metabolites provided putative evidence that they are associated with a specific biosynthetic pathway in Cannabis. The putative pathway proposed provides new insights into the impacts of domestication and breeding pressures for increasing cannabinoid production in strains. The possible overexpression of DXR in the MEP pathway results in higher production of GPP causing the biosynthesis of terpinolene to switch on, which is likely a stress response in the strains to deal with the high concentrations of metabolites (Mahmoud & Croteau, 2001). Seeds are commonly produced by crossing available male plants with desirable traits with known female strains which impacts the expression of terpenes while cannabinoids may be maintained due to male plant selection (Clarke & Merlin, 175 2016b; Soler et al., 2017). When developing “new” strains, breeders select the “ideal” candidate by evaluating the aroma, yield and cannabinoid content or potency (Clarke & Merlin, 2016b). In order to maintain genetic expression, plants are propagated through cloning mother plants or tissue culture (Clarke & Merlin, 2016b; Small, 2015; Wang et al., 2009; Chandra et al., 2017). Due to variable propagation methods utilized by producers, either using seeds, plant cuttings or tissue culture, the phytochemical expression of the strain can vary depending on the genetic expression of the plant grown, which would explain the significant variance that has been observed between strains with the same name through chemical and genetic studies (Sawler et al., 2015; Solar et al., 2017; Hazekamp & Fischedick, 2012; Elzinga et al., 2015). The unique pathway identified can provide insight on several fronts. Firstly, it indicates that some strains do not have this pathway or switch present in its genetic code. There are many very high THC strains that do not produce terpinolene. This could be due to the lack of the terpene synthase for cis-β-ocimene, or because there is not an accumulation of GPP in those strains. Secondly, the artificially increased levels of THC may have invoked the production of monoterpenes that would not previously have been found in natural settings. Cannabis plants in nature, such as wild escapes, typically produce low cannabinoid concentrations at least 10 times lower than seen in today’s markets, therefore the switch to produce terpinolene would not be activated under normal, wild conditions (Clarke & Merlin, 2016b; Small, 2015; ElSohly et al., 2016). Thirdly, the pharmacological effects of terpinolene and its correlated monoterpenes may have played a role in plant selection rather than just aroma. With eight different terpenes strongly correlated to terpinolene, there is a possibility of synergistic effects that would not occur in other strains (Russo, 2011; Lewis et al., 2018; Russo, 2019). The selection of Cannabis strains has focused on many factors including potency, aroma, yield, trichome density and plant size to name a few (Clarke & Merlin, 2016b; Small, 176 2015). Many breeding programs took place in underground operations without access to laboratory equipment and rarely grew sufficient replicates for scientifically valid confirmation of a new cultivar (Clarke & Merlin, 2016b; Small, 2015). This data identified considerable phytochemical overlap between strains indicating that the current system of classifying strains or considering them different does not provide sufficient information to understand the plant diversity. With increasing awareness and interest in delving into the phytochemical diversity of Cannabis, it is necessary to focus on additional metabolites that are present as opposed to just the most abundant. These targeted metabolomic methodologies are just the beginning and are showing unique and interesting information that has not been previously investigated. The use of sensitive and untargeted approaches with advanced chemometric evaluations will continue to improve our understanding of the impacts of breeding, selection and domestication syndrome on Cannabis. The impacts of selection, breeding and domestication have artificially increased the production of major cannabinoids such as total THC and total CBD in many commercially-available strains. With the switching on of the biosynthetic pathways of terpinolene being due to high levels of precursors and cannabinoids within Cannabis, it strongly indicates the impacts of domestication on Cannabis strains.   177 Chapter 7: Conclusion  As of March 31, 2018 there were 296, 702 registered medical marijuana clients in Canada and 15, 618 registered patients for producing their own Cannabis for medical purposes. This encompasses approximately 0.86 % of the Canadian population, with the largest number of patients registered in Ontario (123, 983) and Alberta (112, 207). It is estimated that between 400, 000 and 1, 000, 000 Canadians use Cannabis for therapeutic purposes, but many have relied on purchasing from the illicit market (Walsh et al., 2013). With legalization in effect in Canada as of October 2018, the market is expected to continue to grow. The majority of patients are consuming Cannabis for conditions including HIV/AIDS, spinal pain, arthritis, anxiety/depression, multiple sclerosis, chronic pain, cancer, epilepsy, and many others with symptoms such as sleep deprivation, pain, anxiety, depression, appetite/weight, nausea, inflammation, spasms and many more (Walsh et al., 2013). Due to its legal status, rigorous scientific studies on medicinal Cannabis have been limited and medical doctors have been apprehensive to prescribe it in part due to its previous legal status, limited scientific evidence and phytochemical variation (Walsh et al., 2013; Fletcher, 2013; Belle-Isle et al., 2014). The majority of Cannabis research has focused on isolated cannabinoids or with limited phytochemical characterization. There is considerable anecdotal evidence describing the variable pharmacological effects, while the correlations between therapeutic potential and phytochemical diversity have yet to be fully elucidated.  Cannabis breeding has occurred primarily through clandestine breeding programs, while standardized, highly controlled programs to breed elite varieties or cultivars by selection of phytochemical profiles has been limited (de Meijer, 2014). It is estimated that there are several hundred, perhaps thousands of strains of Cannabis currently being cultivated in legal and illegal markets (Small, 2015). It is possible that chemically identical or very closely related plant material is being sold under different names by different producers and there is no clear 178 definition of the concept of a “strain” (Pollio, 2016; McPartland & Guy, 2017). The impacts of breeding and selection of Cannabis has been shown to exhibit signs of domestication syndrome based on phenotypic expression and genetic structure but has yet to be explored from a phytochemical perspective and how this impacts the variation of Cannabis strains (Sawler et al., 2015; Small, 2015; Clarke & Merlin, 2016b; Soler et al., 2017). The motivation of this project was to explore the impacts of breeding and selection of Cannabis strains on phytochemical diversity, to explore underlying phytochemical relationships using chemometrics and metabolomics and how these may impact human health. With 24, 573 kilograms of Cannabis sold for medical purposes during the 2017-2018 fiscal year in Canada it is evident there is a need for more rigorous understanding of Cannabis phytochemistry. While THC is one of the most studied cannabinoids in Cannabis, there are many negative side effects from taking it as an isolated drug (McPartland et al., 2015). Alternatively, there have been positive correlations with additional metabolites (cannabinoids and/or terpenes) on reducing these undesirable side effects as well as anecdotal evidence suggesting that Cannabis strains with the same THC content can have considerably different pharmacological effects (Russo, & Guy, 2006; Russo, 2011; Hazekamp & Fischedick, 2012l McPartland et al., 2015; Lewis et al., 2018; Russo, 2019). In this case, the evaluation of phytochemical diversity and profiling is necessary for quality control, pre-clinical, clinical and research purposes. Cannabinoids are the most abundant metabolites in Cannabis, which can be upwards of 25 % of the dried flowers by weight. They are concentrated in the glandular trichomes on the female inflorescences. In Canada, quality control regulations only require the quantitation of four cannabinoids: THCA, THC, CBDA and CBD. There have been 120 different cannabinoids characterized in Cannabis, and several cannabinoid standards are commercially-available, therefore methods should include as many cannabinoids as possible 179 (ElSohly et al., 2014). There are a limited number of analytical methods for the quantitation of cannabinoids in Cannabis flowers and many lacked sufficient optimization and validation data to confirm the methods were fit-for-purpose. Additionally, many of the methods relied on environmentally toxic solvents and lacked baseline separation of many cannabinoids. In order to address these issues, a statistically-guided approach using partial factorial designs was undertaken to optimize the extraction of cannabinoids from dried flowers to reduce analysis time, reduce environmental impacts and improve cannabinoid chromatographic separation. The method was validated according to AOAC International guidelines for several performance characteristics including selectivity, precision, accuracy, analyte stability, and limits of detection and quantitation. The final method developed used 80:20 methanol:water %v/v to extract cannabinoids in 15 minutes followed by dilution and a 15 minute chromatographic separation. Validation was completed for the quantitation of ten cannabinoids that were commercially-available at the time of the work to allow for profiling and characterization of Cannabis strains. To evaluate the THC and CBD levels, cannabinoid profiles and the impacts of domestication on Cannabis strains, 33 strains were purchased from 5 different licensed producers throughout Canada. The cannabinoid profiles were characterized for each strain according to the previously validated method for which the content of 32 cannabinoids were obtained where 11 cannabinoids were identified against reference standards and the remaining 21 were unidentified and labeled by elution order as CMPD#. The major cannabinoid contents were used to identify 5 clusters of strains, where the biosynthetic expression of different known cannabinoids based on their precursors olivetolic acid and divarinolic acid were evaluated. Using targeted-untargeted metabolomics it was possible to identify relationships between cannabinoids and putatively identify those correlated with CBDA based on their elution order and UV spectra. The data provided evidence of losses of 180 phytochemical diversity as impacted by domestication and breeding including the loss of CBDA synthase activity in over two thirds of the strains collected. This also impacted the additional CBDA-correlated cannabinoids that have limited known pharmacological significance. There were fewer unidentified cannabinoids in the strains selected for higher THC content. With such strong emphasis on the synthesis of a single metabolite there is a possibility that other biosynthetic pathways have been lost in the process. Ultimately, the data confirm that many of the strains in Canada are closely related based on cannabinoid profiles and current cannabinoid characterization requirements do not provide sufficient information to understand the phytochemical diversity available. Terpenes are the volatile constituents that also accumulate in the glandular trichomes of the female inflorescences. They are considered responsible for the characteristic aromas of Cannabis which has been used to estimate quality and potency in breeding, selection and preference. To evaluate the terpene profiles, the 33 strains were subjected to headspace GC-MS analysis. A total of 67 different terpenes were detected which were grouped according to their cannabinoid clusters to identify relationships between major cannabinoid content and terpene composition within the strains. A total of 5 groups were established. The first group composed of terpenes detected across each cannabinoid class, while the remaining four identified terpenes detected primarily within unique cannabinoid classes. Aromatic characteristics of the terpenes were identified and it was observed that there were some unique aromas that corresponded to different cannabinoid classes, suggesting that there are underlying aromatic notes that breeders used to estimate THC content and/or the presence of CBD when breeding. In addition to these terpene groups, unsupervised metabolomics identified several correlated monoterpenes which were present only in two cannabinoid classes, which further confirms the loss of phytochemical diversity based on terpene composition resulting in domestication syndrome. 181 Cannabinoids and monoterpenes both utilize the precursor geranyl pyrophosphate to produce their final metabolites in the glandular trichomes, therefore it was hypothesized that there may be underlying relationships between these metabolite classes. A multi-block data fusion model was employed to explore the underlying relationships between cannabinoids, monoterpenes and sesquiterpenes from the 33 Cannabis strains. The data identified several unique correlations between terpinolene and a few low abundance cannabinoids indicating the possibility of a unique biosynthetic pathway based on the presence of unique precursors. Other plant species have been shown to produce monoterpenes from neryl pyrophosphate (NPP), an isomer of GPP, which produced many similar monoterpenes to those correlated to terpinolene. Lot-to-lot variance was also evaluated to confirm the consistency between lots and verify these findings. To investigate the relationship between cannabinoids and monoterpenes to further understand this potential biosynthetic link, CMPD12 was isolated from the strain can21. The cannabinoid was identified as THCA-C4 by structure elucidation with NMR and LC-MS analysis. Another cannabinoid that was correlated with these monoterpenes was identified as THCVA using a commercially-available reference standard. While this cannabinoid does not fit with the original hypothesis that this cannabinoid (CMPD12) would be cis-CBGA, it does present valuable information into the potential biosynthetic pathway, suggesting that there is an increased production of precursors present causing the switch of terpinolene production to take place. This is shown with the increased abundance of low-level cannabinoids and the fact that these monoterpenes were only detected in very high THC strains or those with high THC/high CBD strains which have the highest abundance of total cannabinoids. Domestication syndrome can be indicated by the artificial increase in secondary metabolites in plants, for which is has been established that cannabinoid content drops significantly if plants are allowed to propagated in the wild (Small, 2015), therefore breeding has forced the 182 increased production of cannabinoids, which in turn has enabled this terpinolene switch that would otherwise not occur in wild populations. Together, this research provides novel insight into the impacts of breeding, selection and domestication syndrome in Cannabis. There is a significant loss of phytochemical diversity in many strains, which may have impacts on the pharmacological activities of the strains. In order to fully understand the impacts of phytochemical diversity, it is necessary to incorporate sensitive, robust methods for chemical profiling of strains used in medical studies including pre-clinical and clinical evaluations of the effects of Cannabis. Ultimately the goal would be to evaluate the strains using untargeted metabolomics with pharmacological data of the complex materials to develop activity-based classifications that could be used by patients, medical practitioners and the Cannabis industry to better provide recommendations for strain selection for specific medical conditions. Additionally, the use of untargeted metabolomics could provide additional information into the impacts of breeding, selection and domestication syndrome on different metabolites present in Cannabis. 183 References Abdi H, Williams LJ, Valentin D. 2013. Multiple factor analysis: principal component analysis for multitable and multiblock datasets. WIRES Computational Statistics. 5(2): 1519-1535. Adal AM, Sarker LS, Lemke AD, Mahmoud SS. 2017. Isolation and functional characterization of a methyl jasmonate responsive 3-carene synthase from Lavandula x intermedia. Plant Molecular Biology. 93: 641-657. Aizpurua-Olaizola O, Soydaner U, Öztürk E, Schibano D, Simsir Y, Navarro P, Etxebarria N, Usobiaga A. 2016. Evolution of the cannabinoid and terpene content during the growth of Cannabis sativa plants from different chemotypes. Journal of Natural Products. 79(2): 324-331. Alfonsi K, Colberg J, Dunn PJ, Fevig T, Jennings S, Johnson TA, Kleine HP, Knight C, Nagy MA, Perry DA, Stefaniak M. 2008. Green chemistry tools to influence a medicinal chemistry and research chemistry based organization. Green Chemistry. 10: 31-36. Almeida JR, Souza GR, Silva JC, Saraiva SR, Júnior RG, Quintans JD, Barreto RD, Bonjardim LR, Cavalcanti SC, Junior LJ. 2013. Borneol, a bicyclic monoterpene alcohol, reduces nociceptive behavior and inflammatory response in mice. The Scientific World Journal. 2013. 808460. Al-Snafi AE. 2017. The pharmacological and therapeutic importance of Eucalyptus species grown in Iraq. IOSR Journal of Pharmacy. 7(3): 72-91. Amada N, Yamasaki Y, Williams CM, Whalley BJ. 2013. Cannabidivarin (CBDV) suppresses pentylenetetrazole (PTZ)-induced increases in epilepsy-related gene expression. PeerJ. DOI 10.7717/peerj.214 AOAC International. 2013. Appendix K. Part I. AOAC Guidelines for Single-laboratory validation of chemical methods for dietary supplements and botanicals. Official Methods of Analysis. Gaithersburg MD. 184 Babushok VI, Linstom PJ, Zenkevich IG. 2011. Retention indices for frequently reported compounds of plant essential oils. Journal of Physical and Chemical Reference Data. 40(4): 043101. Bäckström B, Cole MD, Carrott MJ, Jones DC, Davidson G, Coleman K. 1997. A preliminary study of the analysis of Cannabis by super critical fluid chromatography with atmospheric pressure chemical ionization mass spectroscopic detection. Science & Justice. 37: 91-97. Bahi A, Mansouri SA, Memari EA, Ameri MA, Nurulain SM, Ojha S. 2014. β-caryophyllene, a CB2 receptor agonist produces multiple behavioral changes relevant to anxiety and depression in mice. Physiology & Behavior. 135: 119-124. Bariyah SK, Ahmed D, Ikram M. 2012. Ocimum Basilicum: A review on phytochemical and pharmacological studies. Pakistan Journal of Chemistry. 2(2): 78-85. Barrett ML, Scott AM, Evans FJ. 1986. Cannflavin A and B, prenylated flavones from Cannabis sativa. Experientia. 42: 452-453. Belle-Isle L, Walsh Z, Callaway R, Lucas P, Capler R, Kay R, Holtzman S. 2014. Barriers to access for Canadians who use Cannabis for therapeutic purposes. International Journal of Drug Policy. 25: 691-699. Berrueta LA, Alonso-Salces RM, Héberger K. 2007. Supervised pattern recognition in food analysis. Journal of Chromatography A. 1158: 197-214. Berry EM, Mechoulam R. 2002. Tetrahydrocannabinol and endocannabinoids in feeding and appetite. Pharmacology & Therapeutics. 95: 185-190. Blaise BJ, Correia G, Tin A, Young JH, Vergnaud AC, Lewis M, Pearce JTM, Elliott P, Nicholson JK, Holmes E, Ebbels TMD. 2016. Power analysis and sample size determination in metabolic phenotyping. Analytical Chemistry. 88(10): 5179-5188. Boccard J, Rudaz S. 2014. Harnessing the complexity of metabolomic data with chemometrics. Journal of Chemometrics. 28: 1-9. 185 Bohlmann F, Hoffmann E. 1979. Cannabigerol-ähnliche Verbindungen aus Helichrysum umbraculigerum. Phytochemistry. 18: 1371–1374. Bohlmann J, Meyer-Gauen G, Croteau R. 1998. Plant terpenoid synthases: Molecular biology and phylogenetic analysis. PNAS. 95(8): 4126-4133. Bolognini D, Costa B, Maione S, Comelli F, Marini P, Di Marzo V, Parolaro D, Ross RA, Guason LA, Cascio MG, Pertwee RG. 2010. The plant cannabinoid Δ9-tetrahydrocannabivarin can decrease signs of inflammation and inflammatory plain in mice. British Journal of Pharmacology. 160: 677-687. Bonjardim LR, Cunha ES, Guimaraes AG, Santana MF, Oliveira MGB, Serafini MR, Araujo AAS, Antoniolli AR, Cavalcanti SCH, Santos MRV, Quintans-Junior LJ. 2012. Evaluations of the anti-inflammatory and antinociceptive properties of p-cymene in mice. Zeitschrift für Naturforschung C. 67(1-2): 15-21. Booth JK, Page JE, Bohlmann J. 2017. Terpene synthases from Cannabis sativa. PLOS One. 12(3): e0173911. Borelli F, Fasolino I, Romano B, Capasso R, Maiello F, Coppola D, Orlando P, Battista G, Pagano E, Di Marzo V, Izzo AA. 2013. Beneficial effect of the non-psychotropic plant cannabinoid cannabigerol on experimental inflammatory bowel disease. Biochemical Pharmacology. 85(9): 1306-1316. Boyd EM, Sheppard P. 1970. Nutmeg oil and camphene as inhaled expectorants. Archives of Otolaryngology. 92(4): 372-378. Breitmaier E. 2016. Terpenes: Flavors, Fragrances, Pharmaca, Pheromones. Wiley-VCH Verlag GmbH & Co. Morlenbach, Germany. Brenneisen R, EISohly M A. 1988. Chromatographic and spectroscopic profiles of Cannabis of different origins: Part I. Journal of Forensic Sciences. 33(6): 1385-1404. Broadhurst DI, Kell DB. 2006. Statistical strategies for avoiding false discoveries in metabolomics and related experiments. Metabolomics. 2(4): 171-196. 186 Brown PN, Chan M, Betz JM. 2010. Optimization and single-laboratory validation study of a high-performance liquid chromatography (HPLC) method for the determination of phenolic Echinacea constituents. Analytical and Bioanalytical Chemistry. 397(5): 1883-1892. Brown PN, Lister P. 2014. Current initiatives for the validation of analytical methods for botanicals. Current Opinion in Biotechnology. 25: 124-128. Burstein SH. 1999. Determination of cannabinoids in Cannabis products using liquid chromatography-ion trap mass spectrometry. Pharmacology & Therapeutics. 82: 87-96. Cadola L, Broseus J, Esseiva P. 2013. Chemical profiling of different hashish seizures by gas chromatography–mass spectrometry and statistical methodology: A case report. Forensic Science International. 232: e24-e27. Calcabrini A, Stringaro A, Toccacieli L, Meschini S, Marra M, Colone M, Arancia G, Molinari A, Salvatore G, Modello F. 2004. Terpinen-4-ol, the main constituent of Melaleuca Alternifolia (tea tree) oil inhibits the in vitro growth of human melanoma cells. Journal of Investigative Dermatology. 122(2): 349-360. California State Legislation Bill. 2014. Senate Bill No. 566. Industrial Hemp. Accessed October 3, 2018 at http://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=201320140SB566  Cannabis Bureau. 2014. Analytical Monograph Cannabis Flos (flowers/granulated). Version 7.1 Accessed October 3, 2018 at https://www.Cannabisbureau.nl/binaries/Cannabisbureau/ documenten/richtlijnen/2017/12/01/monograph-Cannabis-flos/Monograph+Cannabis+Flos+ Version+7.1+%28November+28%2C+2014%29.pdf Casano S, Grassi G, Martini V, Michelozzi M. 2011. Variations in terpene profiles of different strains of Cannabis sativa. Acta Horticulturae. 925: 115-121. Cascaes MM, Guilhon GMSP, de Aguiar Andrade EH, das Graca Bichara Zoghbi M, da Silva Santos L. 2015. Constituents and pharmacological activities of Myrcia (Myrtaceae): A review of an aromatic and medicinal group of plants. International Journal of Molecular Sciences. 16: 23881-23904. 187 Center for Disease Control and Prevention. Chloroform. In: NIOSH Pocket Guide to Chemical Hazards. National Institute for Occupational Safety and Health. Accessed October 3, 2018 at https://www.cdc.gov/niosh/npg/npgd0127.html   Chan KW. 2014. Validating a gas chromatography-mass spectrometric method and sample classification procedure for Cannabis profiling using cannabinoids from case studies. Australian Journal of Forensic Sciences. 46(4): 424-432. Chandra S, Lata H, ElSohly MA, Walker LA, Potter D. 2017. Cannabis cultivation: Methodological issues for obtaining medical grade product. Epilepsy & Behaviour. 70(B): 302-313. Chavan MJ, Wakte PS, Shinde DB. 2010. Analgesic and anti-inflammatory activity of caryophyllene oxide from Annona squamosa L. bark. Phytomedicine. 17(2):149-151. Chen W, Lu Y, Wu J, Gao M, Xu AWB. 2011. Beta-elemene inhibits melanoma growth and metastasis via suppressing vascular endothelial growth factor-mediated angiogenesis. Cancer Chemotherapy and Pharmacology. 67(4): 799-808. Chiou LC, Ling JY, Chang CC. 1997. Chinese herb constituent β-eudesmol alleviated the electroshock seizures in mice and electrographic seizures in rat hippocampal slices. Neuroscience Letters. 231(3): 171-174. Cho JY, Chang HJ, Lee SK, Kim HJ, Hwang JK, Chun HS. 2007. Amelioration of dextran sulfate sodium-induced colitis in mice by oral administration of β-caryophyllene, a sesquiterpene. Life Sciences. 80(10): 932-939. Choa LK, Hua KF, Hsu HY, Cheng SS, Liu JY, Chang ST. 2005. Study on the anti-inflammatory activity of essential oil from leaves of Cinnamomum osmophloeum. Journal of Agricultural and Food Chemistry. 53: 7274-7278. Choi YH, Kim HK, Hazekamp A, Erkelens C, Lefeber AWM, Verpoorte R. 2004. Metabolomic differentiation of Cannabis sativa cultivars using 1H NMR spectroscopy and principal component analysis. Journal of Natural Products. 76: 953.957. 188 Choi YH, Hazekamp A, Peltenburg-Looman AMG, Frederich M, Erkelens C, Lefeber AWM, Verpoorte R. 2004. NMR assignments of the major cannabinoids and cannaflavonoids isolated from flowers of Cannabis sativa. Phytochemical Analysis. 15: 345-354. Choi YH. Verpoorte R. 2014. Metabolomics: What you see is what you extract. Phytochemical Analysis. 25: 289-290. Ciftci O, Ozdemir I, Tanyildizi S, Yildiz S, Oguzturk H. 2011. Antioxidative effects of curcumin, β-myrcene and 1,8-cineole against 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced oxidative stress in rats liver. Toxicology and Industrial Health. 27(5): 447-453. Ciolina LA, Ranieri TL, Taylor AM. 2018. Commercial Cannabis consumer products part 2: HPLC-DAD quantitative analysis of Cannabis cannabinoids. Forensic Science International. 289: 438-447. Clark MN, Bohm BA. 1979. Flavonoid variation in Cannabis L. Botanical Journal of the Linnean Society. 79: 249-257. Clarke RC. 1981. Marijuana botany: an advanced study: the propagation and breeding of distinctive Cannabis. 2nd Ed. Ronin Publishing. Oakland CA. Clarke RC, Merlin MD. 2013. Cannabis: Evolution and Ethnobotany. University of California Press: Los Angeles, USA. Clarke RC, Merlin MD. 2016a. Cannabis Taxonomy: the ‘sativa’ versus ‘indica’ debate. Herbalgram. 110: 44-49. Clarke RC, Merlin MD. 2016b. Cannabis domestication, breeding history, present-day genetic diversity, and future prospects. Critical Reviews in Plant Science. 35: 293-327. Colorado State Regulations. 2015. Industrial Hemp Regulatory Program. Title 35 Agriculture. C.R.S 35-61-109 Comparini IB, Centini F. 1983. Packed column chromatography, high-resolution gas chromatography and high pressure liquid chromatography in comparison for the analysis of Cannabis constituents. Forensic Science International. 21: 129-137. 189 Consroe P, Martin A, Singh V. 1961. Antiepileptic potential of cannabidiol analogs. Journal of Clinical Pharmacology. 21: 428S-436S. Crawford I, Ruske S, Toping DO, Gallagher MW. Evaluation of hierarchical agglomerative cluster analysis methods for discrimination of primary biological aerosol. Atmospheric Measurement Techniques. 8: 4979-4991. De Albuquerque IL, Alves LA, Lemos TLG, Dorneles CA, de Morais MO. 2008. Constituents of the essential oil of Brazilian green propolis from Brazil. Journal of Essential Oil Research. 20: 414-415. De Backer B, Debrus B, Lebrun P, Theunis L, Dubois N, Decock L, Verstraete P, Hubert P, Charlier C. 2009. Innovative development and validation of an HPLC/DAD method for the qualitative and quantitative determination of major cannabinoids in Cannabis plant material. Journal of Chromatography B. 877: 4115-4124. De Backer B, Maebe K, Verstraete AG, Charlier C. 2012. Evolution of the content of THC and other major cannabinoids in drug-type Cannabis cuttings and seedlings during growth of plants. Journal of Forensic Sciences. 57(4): 918-922. Degenhardt F, Stehle F, Kayser O. 2017. Chapter 2: The biosynthesis of cannabinoids. Handbook of Cannabis and Related Pathologies. Preedy VR, Ed. Elsevier, London, UK. Pp. 12-23. De Meijer EPM, Bagatta M, Carboni A, Crucitti P, Moliterni VMC, Ranalli P, Mandolino G. 2003. The inheritance of chemical phenotype in Cannabis sativa L. Genetics. 163(1): 335-346. De Meijer E. 2014. The chemical phenotypes (Chemotypes) of Cannabis. Handbook of Cannabis. Pertwee, R. (editor). Oxford University Press, p. 89-110. De Meijer EPM, Hammond KM, Micheler M. 2009. The inheritance of chemical phenotype in Cannabis sativa L. (III): variation in cannabichromene proportion. Euphytica. 165(2): 293-311. de Oliveira MG, Marques RB, de Santana MF, Santos AB, Brito FA, Barreto EO, De Sousa DP, Almeida FR, Badauê‐Passos Jr D, Antoniolli ÂR, Quintans‐Júnior LJ. 2012. α‐Terpineol 190 reduces mechanical hypernociception and inflammatory response. Basic & Clinical Pharmacology & Toxicology. 111(2): 120-125. de Oliveira TM, de Carvalho RB, da Costa IH, de Oliveira GA, de Souza AA, de Lima SG, de Freitas RM. 2015. Evaluation of p-cymene, a natural antioxidant. Pharmaceutical Biology. 53(3): 423-428. Devinsky O, Cilio MR, Cross H, Fernandez-Ruiz J, French J, Hill C, Katz R, Di Marzo V, Jutras-Aswad D, Notcutt WG, Martinez-Orgado J, Robson PJ, Rohrback BG, Thiele E, Whalley B, Friedman D. 2014. Cannabidiol: pharmacology and potential therapeutic role in epilepsy and other neuropsychiatric disorders. Epilepsia. 55(6): 791-802. de Sousa DP, Nobrega FFF, de Morais LSCL, de Almeida RN. 2009. Evaluation of the anticonvulsant activity of terpinen-4-ol. Zeitschrift für Naturforschung C. 64: 1-5. Dunn WB, Erban A, Weber RJM, Creek DJ, Brown M, Breitling R, Hankemeier T, Goodacre R, Neumann S, Kopka J, Viant MR. 2012. Mass appeal: Metabolite identification in mass spectrometry-focused untargeted metabolomics. Metabolomics. 9: 44-66. El-Alfy AT, Ivey K, Robinson K, Ahmed S, Radwan M, Slade D, Khan I, ElSohly M, Ross S. 2010. Antidepressant-like effect of Δ9-tetrahydrocannabinol and other cannabinoids isolated from Cannabis sativa L. Pharmacology Biochemistry and Behaviour. 95(4): 434-442. El Hadri A, del Río MÁ, Sanz J, Coloma AG, Idaomar M, Ozonas BR, González JB, Reus MI. 2010. Cytotoxic activity of α-humulene and trans-caryophyllene from Salvia officinalis in animal and human tumor cells. Anales de la Real Academia Nacional de Farmacia. 76(3): 343-356. ElSohly MA, Ross SA, Mehmedic Z, Arafat R, Yi Bao, Banaha BF. 2000. Potency trends of Δ9-THC and other cannabinoids in confiscated marijuana from 1980-1997. Journal of Forensic Sciences. 45(1): 24-30. ElSohly MA, Slade D. 2005. Chemical constituents of marijuana: The complex mixture of natural cannabinoids. Life Sciences. 78: 539-548. 191 ElSohly MS, Stanford DF, Murphy TP. 2007. Chemical fingerprinting of Cannabis as a means of source identification. Forensic Science and Medicine: Marijuana and the Cannabinoids. ElSohly MA Ed. Humana Press Inc. Totowa, New Jersey. Pp. 205-235. ElSohly MA, Gul W. 2014. Constituents of Cannabis sativa. Handbook of Cannabis. Roger Pertwee Ed. Oxford University Press, Oxford UK. p. 3-22. ElSohly MA, Mehmedic Z, Foster S, Gon C, Chandra S, Church JC. 2016. Changes in Cannabis potency over the last 2 decades (1995-2014): Analysis of current data in the United States. Biological Psychiatry. 79(7): 613-619. Elzinga S, Fischedick J, Podkolinski R, Raber JC. 2015. Cannabinoids and terpenes as chemotaxonomic markers in Cannabis. Natural Products Chemistry & Research. 3: 181-190. Environmental Protection Agency. Guidelines establishing test procedures for the analysis of pollutants; Procedures for detection and quantification. In: 40 CFR pt. 136, Appendix D, rev. 1.11. 2002. Accessed October 3, 2018 at https://www.gpo.gov/fdsys/pkg/CFR-2011-title40-vol23/pdf/CFR-2011-title40-vol23-part136-appD.pdf EU Commission. 2000. Commission Regulation (EC) No 2860/2000. Official Journal of the European Communities. L332/63. Accessed October 3, 2018 at http://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32000R2860&from=EN Fellermeier M, Zenk MH. 1998. Prenylation of olivetolate by a hemp transferase yields cannabigerolic acid, the precursor of tetrahydrocannabinol. FEBS Letters. 427: 283–285 Fernandes ES, Passos GF, Medeiros R, da Cunha FM, Ferreira J, Campos MM, Pianowski LF, Calixto JB. 2007. Anti-inflammatory effects of compounds alpha-humulene and (−)-trans-caryophyllene isolated from the essential oil of Cordia verbenacea. European Journal of Pharmacology. 569(3): 228-236. Fetterman PS, Doorenbos NJ, Keith ES, Quimey MW. 1971. A simple gas liquid chromatography procedure for determination of cannabinoidic acids in Cannabis sativa L. Experientia. 27(8): 988-990. 192 Fiehn O. 2002. Metabolomics – the link between genotypes and phenotypes. Plant Molecular Biology. 48: 155-171. Fischedick JT, Hazekamp A, Erkelens T, Choi YH, Verpoorte R. 2010. Metabolic fingerprinting of Cannabis sativa L., cannabinoids and terpenoids for chemotaxonomic and drug standardization purposes. Phytochemistry. 71: 2058-2073. Fish F. 1974. Gas chromatography of Cannabis constituents. Chromatographia. 7(6): 311-315. Fletcher J. 2013. Marijuana is not a prescription medicine. Canadian Medical Association Journal. 185(5): 369. Flores-Sanchez IJ, Verpoorte R. 2008. Secondary metabolism in Cannabis. Phytochemical Reviews. 7: 615-639. Fournier G, Richez-Dumanois C, Duvezin J, Mathieu JP, Paris M. 1987. Identification of a new chemotype in Cannabis sativa: cannabigerol-dominant plants, biogenetic and agronomic prospects. Planta Medica. 53: 277–280. Freitas JC, Presgrave OA, Fingola FF, Menezes MA, Paumgartten FJ. 1993. Effect of beta-myrcene on pentobarbital sleeping time. Brazilian Journal of Medical and Biological Research. 26(5): 519-523. Fu Z, Wang H, Hu X, Sun Z, Han C. 2013. The pharmacological properties of salvia essential oils. Journal of Applied Pharmaceutical Science. 3(7): 122-127. Galdino PM, Nascimento MV, Florentino IF, Lino RC, Fajemiroye JO, Chaibub BA, de Paula JR, de Lima TC, Costa EA. 2012. The anxiolytic-like effect of an essential oil derived from Spiranthera odoratissima A. St. Hil. leaves and its major component, β-caryophyllene, in male mice. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 38(2): 276-284. Gaoni Y, Mechoulam R. 1964. Isolation, structure and partial synthesis of an active constituent in hashish. Journal of the American Chemistry Society. 86(8): 1646-1647. 193 Garg P, Sardana S. 2016. Pharmacological and therapeutic effects of Ocimum sanctum. European Journal of Pharmaceutical and Medical Research. 3(8): 637-640. Giese MW, Lewis MA, Giese L, Smith KM. 2015. Development and validation of a reliable and robust method for the analysis of cannabinoids in terpenes in Cannabis. Journal of AOAC International. 98(6): 1503-1522. Gilbert AN, Di Verdi JA. 2018. Consumer perceptions of strain differences in Cannabis aroma. PLOS One. 13(2): e0192247.  Giwanon R, Thubthimthed S, Rerk-am U, Sunthorntanasart T. 2000. Antimicrobial activity of terpinen-4-ol and sabinene. Thai Journal of Pharmaceutical Sciences. 24: 27. Gul W, Gul SW, Radwan MM, Wanas AS, Mehmedic Z, Khan IA, Sharaf MHM, ElSohly MA. 2015. Determination of 11 cannabinoids in biomass and extracts of different varieties of Cannabis using high-performance liquid chromatography. Journal of AOAC International. 98(6): 1523-1528. Gurgel do Vale T, Couto Furtado E, Santos JG, Viana GSB. 2002. Central effects of citral, myrcene and limonene, constituents of essential oil chemotypes from Lippia alba (Mill.) N.E. Brown. Phytomedicine. 9(8): 709-714. Guzman-Gutierrez SL, Bonilla-Jamie H, Gomez-Cansino R, Reyes-Chilpa R. 2015. Linalool and β-pinene exert their antidepressant-like activity through the monoaminergic pathway. Life Sciences. 128: 24-29. Hadi A, Hossein N, Shirin P, Majmeh N, Abolfazi M. 2014. Anti-inflammatory and analgesic activities of Artemisia absinthium and chemical composition of its essential oil. International Journal of Pharmaceutical Sciences Review and Research. 38: 237-244. Hall RD. 2011. Plant metabolomics in a nutshell: Potential and future challenges. Annual Plant Reviews. 43: 1-24. Hartmann T. 2007. From waste products to ecochemicals: Fifty years of research of plant secondary metabolism. Phytochemistry. 68(22-24): 2831-2846. 194 Harvey DI. 1976. Characterization of the butyl homologues of ∆1-tetrahydrocannabinol, cannabinol and cannabidiol in samples of Cannabis by combined gas chromatography and mass spectrometry. Journal of Pharmacy and Pharmacology. 28(4): 280-285. Hazekamp A, Choi YH, Verpoorte R. 2004. Quantitative analysis of cannabinoids from Cannabis sativa using 1H-NMR. Chemical and Pharmaceutical Bulletin. 52(6): 718-721. Hazekamp A, Fischedick JT. 2012. Cannabis – from cultivar to chemovar. Drug Testing and Analysis. 4: 660-667. Hazekamp A, Tejkalova K, Papadimitriou S. 2016. Cannabis: From cultivar to chemovar II: A metabolomics approach to Cannabis classification. Cannabis and Cannabinoid Research. 1(1): 202-215. Health Canada. 1998. Industrial Hemp Regulations. SOR/98-156. Health Canada. 2016. Access to Cannabis for Medical Purposes Regulations. SOR/2016-230. Health Canada. 2018. Cannabis Regulations. SOR/2018-144. Hegnauer R. 1986. Phytochemistry and plant taxonomy – An essay on the chemotaxonomy of higher plants. Phytochemistry. 25(7): 1519-1535. Hieu LD, Hoi TM, Thang TD, Eresanya OI, Ogunwande IA. 2018. Fonenol, the main constituent of the essention oil of the leaf of Piper longum L. American Journal of Essential Oils and Natural Products. 6(1): 16-19. Hill AJ, Weston Se, Jones NA, Smith I, Bevan SA, Williamson EM, Stephens GJ, Williams CM, Whalley BJ. 2010. Δ9-tetrahydrocannabivarin suppresses in vitro epileptiform and in vivo seizure activity in rats. Epilepsia. 51(8): 1522-1532. Hillig K. 2004. A chemotaxonomic analysis of terpenoid variation in Cannabis. Biochemical Systematics and Ecology. 32: 875-891. Hillig K, Mahlberg PG. 2004. A chemotaxonomic analysis of cannabinoid variation in Cannabis (Cannabaceae). American Journal of Botany. 91(6): 966-975. 195 Hillig KW. 2005. Genetic evidence for speciation of Cannabis (Cannabaceae). Genetic Resources and Crop Evolution. 52(2): 161-180. Him A, Ozbek H, Turel I, Oner AC. 2008. Antinociceptive activity of alpha-pinene and fenchone. Pharmacologyonline. 3: 363-369. Horwitz W. 1982. Evaluation of analytical methods used for regulation of foods and drugs. Analytical Chemistry. 54(1): 67-76. Hsu HC, Yang WC, Tsai WJ, Chen CC, Huang HY, Tsai YC. 2006. α-Bulnesene, a novel PAF receptor antagonist isolated from Pogostemon cablin. Biochemical and Biophysical Research Communications. 345(3):1033-1038. Huestis MA, Smith ML. 2007. Human cannabinoid pharmacokinetics and interpretation of cannabinoid concentrations in biological fluids and tissues. Chapter 9. Forensic Science and Medicine: Marijuana and the Cannabinoids. ElSohly MA Ed. Humana Press Inc. Totowa, New Jersey. Pp. 205-235. Ilias Y, Rudaz S, Mathieu P, Christen P, Veuthey JL. 2005. Extraction and analysis of different Cannabis samples by headspace solid-phase microextraction combined with gas chromatography-mass spectrometry. Journal of Separation Science. 28: 2293-2300. Inui T, Wang Y, Pro SM, Franzblau SG, Pauli GF. 2012. Unbiased evaluation of bioactive secondary metabolites in complex matrices. Fitoterapia. 83(7): 1218-1225. Isahq MS, Afridi MS, Ali J, Hussain MM, Ahmad S, Kanwal F. 2015. Proximate composition, phytochemical screening, GC-MS studies of biologically active cannabinoids and antimicrobial activities of Cannabis indica. Asian Pacific Journal of Tropical Disease. 5(11): 897-902. Ito K, Ito M. 2013. The sedative effect of inhaled terpinolene in mice and its structure-activity relationships. Journal of Natural Medicines. 67(4): 833-837. Izzo AA, Capasso R, Aviello G, Borrelli F, Romano B, Piscitelli F, Gallo L, Capasso F, Orlando P, Di Marzo V. 2012. Inhibitory effect of cannabichromene, a major non-psychotropic cannabinoid extracted from Cannabis sativa, on inflammation-induced hypermotility in mice. British Journal of Pharmacology. 166: 1444-1460. 196 Janischevsky DE. 1924. Cannabis ruderalis. Proceedings of Saratov University. 2: 14-15. Jin D, Jin S, Yu Y, Lee C, Chen J. 2017. Classification of Cannabis cultivars marketed in Canada for medical purposes by quantification of cannabinoids and terpenes using HPLC-DAD and GC-MS. Journal of Analytical & Bioanalytical Techniques. 8(1): 349. Kanter SL, Musumeci MR, Hollister LE. 1979. Quantitative determination of Δ9-tetrahydrocannabinol and Δ9-tetrahydrocannabinolic acid in marihuana by high-pressure liquid chromatography. Journal of Chromatography A. 171: 504-508. Katsuyama S, Mizoguchi H, Kuwahata H, Komatsu T, Nagaoka K, Nakamura H, Bagetta G, Sakurada T, Sakurada S. 2013. Involvement of peripheral cannabinoid and opioid receptors in β‐caryophyllene‐induced antinociception. European Journal of Pain. 17(5): 664-675. Kim DS, Lee HJ, Jeon YD, Han YH, Kee JY, Kim HJ, Shin HJ, Kang JW, Lee BS, Kim SH, Kim SJ, Park SH, Choi BM, Park SJ, Um JY, Hong SH. 2015. Alpha-pinene exhibits anti-inflammatory activity through the suppression of MAPKs and NF-kB pathway in mouse peritoneal macrophages. The American Journal of Chinese Medicine. 43(4): 731-742. Kimura IK, Tsuneki HI, Kondoh TA, Kimura MA. 1991. Enhancing effect by nicotinic acetylcholine receptor channel blockers, including beta-eudesmol, on succinylcholine-induced inhibition of twitch tension and intracellular Ca++ in mouse diaphragm muscle. Journal of Pharmacology and Experimental Therapeutics. 256(1): 24-28. Kimura Y, Sumiyoshi M. 2012. Effects of an Atractylodes lancea rhizome extract and a volatile component β-eudesmol on gastrointestinal motility in mice. Journal of Ethnopharmacology. 141(1): 530-536. Klein TW. 2005. Cannabinoid-based drugs as anti-inflammatory therapeutics. Nature Reviews Immunology. 5: 400-411. Kuroda K, Inoue N, Ito Y, Kubota K, Sugimoto A, Kakuda T, Fushiki T. 2005. Sedative effects of the jasmine tea odor and (R)-(−)-linalool, one of its major odor components, on autonomic nerve activity and mood states. European Journal of Applied Physiology. 95: 107-114. 197 Lahlou S, Interaminense LF, Leal‐Cardoso JH, Duarte GP. 2003. Antihypertensive effects of the essential oil of Alpinia zerumbet and its main constituent, terpinen‐4‐ol, in DOCA‐salt hypertensive conscious rats. Fundamental & Cinical Pharmacology. 17(3): 323-330. Lamarine RJ. 2012. Marijuana: Modern medical chimaera. Journal of Drug Education. 42(1): 1-11. Lehmann T, Brenneisen. 1995. High performance liquid chromatographic profiling of Cannabis products. Journal of Liquid Chromatography. 18(4): 689-700. Leonhardt V, Leal‐Cardoso JH, Lahlou S, Albuquerque AA, Porto RS, Celedônio NR, Oliveira AC, Pereira RF, Silva LP, Garcia‐Teófilo TM, Silva AP. 2010. Antispasmodic effects of essential oil of Pterodon polygalaeflorus and its main constituent β‐caryophyllene on rat isolated ileum. Fundamental & Clinical Pharmacology. 24(6): 749-758. Lewis MA, Russo EB, Smith KM. 2018. Pharmacological foundations of Cannabis chemovars. Planta Medica. 84: 225-233. Lima DF, Brandao MS, Moura JB, Leitao JMRS, Carvalho FAA, Miura LMCV, Leite JRSA, Sousa DP, Almeida FRC. 2012. Antinociceptive activity of the monoterpene α‐phellandrene in rodents: possible mechanisms of action. Journal of Pharmacy and Pharmacology. 64(2): 283-292. Linck VM, da Silva AL, Figueiro M, Caramao Eb, Moreno PRH, Elisabetsky E. 2010. Effects of inhaled linalool in anxiety, social interaction and aggressive behavior in mice. Phytomedicine. 17: 679-683. Maa E, Figi P. 2014. The case for medical marijuana in epilepsy. Epilepsia. 55(6): 783-786. Macedo EMA, Santos WC, Sousa Neto BP, Lopes EM, Piauilino CA, Cunha FVM, Sousa DP, Oliveira FA, Almeida FRC. 2016. Association of terpinolene and diclofenac presents antinociceptive and anti-inflammatory synergistic effects in a model of chronic inflammation. Brazilian Journal of Medical and Biological Research. 49(7): e5103. Mahlberg PG, Kim ES. 2004. Accumulation of cannabinoids in glandular trichomes of Cannabis (Cannabaceae). Journal of Industrial Hemp. 9(1): 15-36. 198 Mahmoud SS, Croteau RB. 2001. Metabolic engineering of essential oil yield and composition in mint by altering expression of deoxyxylulose phosphate reductoisomerase and menthofuran synthase. PNAS. 98(15): 8915-8920. McKey D, Elias M, Pujol D, Duputie A. 2010. The evolutionary ecology of clonally propagated domesticated plants. New Phytologist. 186: 318-332. McLaren J, Swift W, Dillon P, Allsop S. 2008. Cannabis potency and contamination: a review of the literature. Addiction. 103: 1100-1109. McPartland JM. 2018. Cannabis systematics at the level of family, genus and species. Cannabis and Cannabinoid Chemistry. 3(1): 203-212. McPartland JM, Russo EB. 2001. Cannabis and Cannabis extracts: Greater than the sum of their parts? Journal of Cannabis Therapeutics. 1(2-3): 103-132. McPartland JM, Duncan M, Di Marzo V, Pertwee RG. 2015. Are cannabidiol and Δ9-tetrahydrocannabivarin negative modulators of the endocannabinoid system? A systemic review. British Journal of Pharmacology. 172: 737-753. McPartland JM, Guy GW. 2017. Models of Cannabis taxonomy, cultural bias and conflicts between scientific and vernacular names. The Botanical Review. 83: 327-381. Mechoulam R, Peters M, Murillo-Rodriguez E, Hanus LO. 2007. Cannabidiol – recent advances. Chemistry and Biodiversity. 4: 1678-1692. Mediavilla V, Steinemann S. 1997. Essential oil of Cannabis sativa L. strains. Journal of the International Hemp Association. 4(2): 80-82. Mehmedic Z, Chandra S, Slade D, Denham H, Foster S, Patel AS, Ross SA, Khan IA, ElSohly MA. 2010. Potency trends of Δ9-THC and other cannabinoids in confiscated Cannabis preparations from 1993 to 2008. Journal of Forensic Sciences. 55(5): 1209-1217. Meinicke P, Lingner T, Kaever A, Feussner K, Göbel C, Feussner I, Karlovsky P, Morgenstern B. 2008. Metabolite-based clustering and visualization of mass spectrometry data using one-dimensional self-organizing maps. Algorithms for Molecular Biology. 3: 9. 199 Meyer RS, DuVal AE, Jensen HR. 2012. Patterns and processing in crop domestication: a historical review and quantitative analysis of 203 global food crops. New Phytologist. 196: 299-348. Milde J, Elstner EF, Grassmann J. 2004. Synergistic inhibition of low-density lipoprotein oxidation by rutin, γ-terpinene, and ascorbic acid. Phytomedicine. 11: 105-113. Miller JA, Lang JE, Ley M, Nagle RB, Hsu CH, Thompson P, Cordova C, Waer A, Chow HS. 2013. Human breast tissue disposition and bioactivity of limonene in women with early stage breast cancer. Cancer Prevention Research. 6(6): 577-584 Morimoto S, Tanaka Y, Shoyama Y. 1999. Biosynthesis of cannabinoids in Cannabis sativa L. Current Topics in Phytochemistry. 2: 103-113. Mudge EM, Paley L, Schieber A, Brown PN. 2015a. Optimization and single-laboratory validation of a method for the determination of flavonolignans in milk thistle seeds by high-performance liquid chromatography with ultraviolet detection. Analytical and Bioanalytical Chemistry. 407(25): 7657-7666. Mudge EM, Jones AMP, Brown PN. 2015b. Quantification of pyrrolizidine alkaloids in North American plants and honey by LC-MS: single laboratory validation. Food Additives and Contaminants. 32(12): 2068-2077. Mudge EM, Betz JM, Brown PN. 2016a. The importance of method selection in determining product integrity for nutrition research.  Advances in Nutrition. 7: 390-398. Mudge EM, Chan M, Venkataraman S, Brown PN. 2016b. Curcuminoids in turmeric roots and supplements: Method optimization and validation. Food Analytical Methods. 9: 1428-1435. Mudge EM, Liu Y, Lund JA, Brown PN. 2016c. Single-laboratory validation for the determination of flavonoids in hawthorn leaves and finished products by LC-UV. Planta Medica. 82(17): 1487-1492. Mudge EM, Brown PN. 2017. Determination of mitragynine in Mitragyna speciosa raw materials and finished products by liquid chromatography with UV detection: Single-laboratory validation. Journal of AOAC International. 100(1): 18-24. 200 Mukaka MM. 2012. A guide to appropriate use of correlation coefficient in medical research. Malawi Medical Journal. 24(3): 69-71. Murali R, Karthikeyan A, Saravanan R. 2013. Protective effects of D‐limonene on lipid peroxidation and antioxidant enzymes in streptozotocin‐induced diabetic rats. Basic & Clinical Pharmacology & Toxicology. 112(3): 175-181. Nam JH, Nam DY, Lee DU. 2016. Valencene from the rhizomes of Cyperus rotundus inhibits skin photoaging-related ion channels and UV-induced melanogenesis in B16F10 melanoma cells. Journal of Natural Products. 79(4): 1091-1096. New York State Register. Chapter XIII Medical Use of Marihuana. Part 1004. Commissioner of Health by Section 3369-a. Accessed July 12, 2016 at https://www.health.ny.gov/ regulations/medical_marijuana/docs/regulations.pdf. Nicholson JK, Connelly H, Lindon JC, Holmes E. 2002. Metabonomics: a platform for studying drug toxicity and gene function. Nature Reviews. 1: 153-161. Nigam MC, Handa KL, Nigam IC, Levi L. 1965. Essential oils and their constituents XXIX. The essential oil of marihuana: composition of genuine Indian Cannabis sativa L. Canadian Journal of Chemistry. 43: 3372-3376. Noestheden M, Friedlander G, Anspach J, Krepich S, Hyland KC, Zandberg WF. 2018. Chromatographic characterization of 11 phytocannabinoids: Quantitative and fit-to-purpose performance as a function of extra-column variance. Phytochemical Analysis. 29(5): 507-515. Ohara K, Fukuda T, Ishida Y, Takahashi C, Ohya R, Katayama M, Uchida K, Tominaga M, Nagai K. 2017. β-Eudesmol, an oxygenized sesquiterpene, stimulates appetite via TRPA1 and the autonomic nervous system. Scientific Reports. 7(1): 15785. Orellana-Paucar AM, Serruys ASK, Afrikanova T, Maes J, De Borggraeve W, Alen J, Leon-Tamariz F, Wilches-Arizabala IM, Crawford AD, de Witte PAM, Esguerra CV. 2012. Anticonvulsant activity of bisabolene sesquiterpenoids of Curcuma longa in zebrafish and mouse seizure models. Epilepsy & Behavior. 24: 14-22. 201 Pang Y, Wang D, Fan Z, Chen X, Yu F, Hu X, Wang K, Yuan L. 2014. Blumea balsamifera – a phytochemical and pharmacological review. Molecules. 19: 9453-9477. Park KR, Nam D, Yun HM, Lee SG, Jang HJ, Sethi G, Cho SK, Ahn KS. 2011. β-Caryophyllene oxide inhibits growth and induces apoptosis through the suppression of PI3K/AKT/mTOR/S6K1 pathways and ROS-mediated MAPKs activation. Cancer Letters. 312(2): 178-188. Passos FF, Lopes EM, de Araújo JM, de Sousa DP, Veras LM, Leite JR, Almeida FR. 2015. Involvement of cholinergic and opioid system in γ-terpinene-mediated antinociception. Evidence-Based Complementary and Alternative Medicine. 2015: 829414. Patel B, Wene D, Fan Z. 2017. Qualitative and quantitative measurements of cannabinoids in Cannabis using modified HPLC/DAD method. Journal of Pharmaceutical and Biomedical Analysis. 146: 15-23. Paumgartten FJR, Delgado IF, Alves EN, Nogueira ACMA, De-Farias RC, Neubert D. 1990. Single dose toxicity study of β-myrcene, a natural analgesic substance. Brazilian Journal of Medical and Biological Research. 23: 873-877. Peana AT, Paolo SD, Chessa ML, Moretti MD, Serra G, Pippia P. 2003. (−)-Linalool produces antinociception in two experimental models of pain. European Journal of Pharmacology. 460(1): 37-41. Peng W, Han T, Xin WB, Zhang XG, Zhang QY, Jia M, Qin LP. 2011. Comparative research of chemical constituents and bioactives between petroleum ether extracts of the aerial part and the rhizome of Atracylodes macrocephala. Medicinal Chemistry Research. 20: 146-151. Pertwee RG. 2008. The diverse CB1 and CB2 receptor pharmacology of three plant cannabinoids: Δ9-tetrahydrocannabinol, cannabidiol and Δ9-tetrahydrocannabivarin. British Journal of Pharmacology. 153: 199-215. Piluzza G, Delogu G, Cabras A, Marceddu S, Bullitta S. 2013. Differentiation between fiber and drug types of hemp (Cannabis sativa L.) from a collection of wild and domesticated accessions. Genetic Resources and Crop Evolution. 60: 2331-2342. 202 Plasse TF, Gorter RW, Krasnow SH, Lane M, Shepard KV, Wadleigh RG. 1991. Recent clinical experience with dronabinol. Phamacology Biochemistry and Behavior. 40(3): 695-700. Plengsuriyakarn T, Karbwang J, Na‐Bangchang K. 2015. Anticancer activity using positron emission tomography‐computed tomography and pharmacokinetics of β‐eudesmol in human cholangiocarcinoma xenografted nude mouse model. Clinical and Experimental Pharmacology and Physiology. 42(3): 293-304. Politi W, Peschel W, Wilson N, Zloh M, Prieto JM, Heinrich M. 2008. Direct NMR analysis of Cannabis water extracts and tinctures and semi-quantitative data on Δ9-THC and Δ9-THC-acid. Phytochemistry. 69: 562-570. Pollio A. 2016. The name of Cannabis: A short guide for nonbotanists. Cannabis and Cannabinoid Research. 1(1): 234-238. Porter BE, Jacobson C. 2013. Report of a parent survey of cannabidiol-enriched Cannabis use in pediatric treatment-resistant epilepsy. Epilepsy & Behavior. 29(3): 574-577. Quintans JD, Menezes PP, Santos MR, Bonjardim LR, Almeida JR, Gelain DP, de Souza Araújo AA, Quintans-Júnior LJ. 2013. Improvement of p-cymene antinociceptive and anti-inflammatory effects by inclusion in β-cyclodextrin. Phytomedicine. 20(5): 436-440. Radwan MM, ElSohly MA, Slade D, Ahmed SA, Wilson L, El-Alfy AT, Khan IA, Ross SA. 2008. Non-cannabinoid constituents from a high potency Cannabis sativa variety. Phytochemistry. 69(14): 2627-2633. Radwan MM, ElSohly MA, El-Alfy AT, Ahmed SA, Slade D, Husni AS, Manle SP, Wilson L, Seale S, Cutler SJ, Ross SA. 2015. Isolation and pharmacological evaluation of minor cannabinoids from high-potency Cannabis sativa. Journal of Natural Products. 78(6): 1271-1276. Raharjo TJ, Verpoorte R. 2004. Methods for the analysis of cannabinoids in biological materials: a review. Phytochemical Analysis. 15: 79-94. 203 Ram HM, Sett R. 1982. Induction of fertile male flowers in genetically female Cannabis sativa plants by silver nitrate and silver thiosulphate anionic complex. Theoretical and Applied Genetics. 62(4): 369-375. Ringrose PS, Parr MA, McLaren M. 1975. Effects of anti-inflammatory and other compounds on the release of lysosomal enzymes from macrophages. Biochemical Pharmacology. 24(5): 607-614. Robson P. 2001. Therapeutic aspects of Cannabis and cannabinoids. British Journal of Psychiatry. 178: 107-115. Rock EM, Goodwin JM, Limebeer CL, Breuer A, Pertwee RG, Mechoulam R, Parker LA. 2011. Interaction between non-psycotropic cannabinoids in marihuana: effect of cannabigerol (CBG) and the anti-nausea or anti-emetic effects of cannabidiol (CBD) in rats and shrews. Psychopharmacology. 215: 505-512. Rock EM, Sticht MA, Duncan M, Stott C, Parker LA. 2013. Evaluation of the potential of the phytocannabinoids, cannabidivarin (CBDV) and Δ9-tetrahydrocannabivarin (THCV), to produce CB1 receptor inverse agonism symptoms of nausea in rats. British Journal of Pharmacology. 170: 671-678. Rogerio AP, Andrade EL, Leite DFP, Figueiredo CP, Calixto JB. Preventative and therapeutic anti-inflammatory properties of the sesquiterpene α-humulene in experimental airways allergic inflammation. British Journal of Pharmacology. 158(4): 1074-1087. Romano LL, Hazekamp A. 2013. Cannabis oil: chemical evaluation of an upcoming Cannabis-based medicine. Cannabinoids. 1(1): 1-11. Ross SA, ElSohly MA. 1996. The volatile oil composition of fresh and air-dried buds of Cannabis sativa. Journal of Natural Products. 59: 49-51. Ruckebusch C, Blanchet L. 2013. Multivariate curve resolution: a review of advanced and tailored applications and challenges. Analytica Chimica Acta. 765: 28-36. Russo EB, Burnett A, Hall B, Parker KK. 2005. Agonistic properties of cannabidiol at 5-HT-1a receptors. Neurochemical Research. 30: 1037–1043. 204 Russo E, Guy GW. 2006. A tale of two cannabinoids: the therapeutic rationale for combining tetrahydrocannabinol and cannabidiol. Medical Hypotheses. 66(2): 234-246. Russo EB, Guy GW, Robson PJ. 2007. Cannabis, pain, and sleep: Lessons from therapeutic clinical trials of Sativex, a Cannabis-based medicine. Chemistry & Biodiversity. 4: 1729-1743. Russo EB. 2011. Taming THC: potential Cannabis synergy and phytocannabinoid-terpenoid entourage effects. British Journal of Pharmacology. 163: 1344-1364. Russo EN. 2019. The case for the entourage effect and conventional breeding of clinical Cannabis: No “strain”, no gain. Frontiers in Plant Science. 9: 1969. Sawler J, Stout JM, Gardner KM, Hudson D, Vidmas J, Butler L, Page JE, Myles S. 2015. The genetic structure of marijuana and hemp. PLoS ONE 10: e0133292. Satou T, Kasuya H, Maeda K, Koike K. 2014. Daily inhalation of α-pinene in mice: effects on behaviou and organ accumulation. Phytotherapy Research. 28(9): 1284-1287. Schilmiller A, Schauvinhold I, Larson M, Xu R, Charbonneau AL, Schmidt A, Wilkerson C, Last RL, Pichersky E. 2009. Monoterpenes in the glandular trichomes of tomato are synthesized from a neryl diphosphate precursor rather than geranyl diphosphate. PNAS. 106(26): 10865-10870. Schultes RE, Klein WM, Plowman T, Lockwood TE. 1975. Cannabis: An example of taxonomic neglect. Cannabis and Culture. Vera Rubin Ed. Mouton Publishers. The Hague, Netherlands. Segelman AB, Segelman FP, Varma SD, Wagner H, Seligmann O. 1977. Cannabis sativa L. (Marijuana) IX: Lens aldose reductase inhibitory activity of marijuana flavone C-glycosides. Journal of Pharmaceutical Sciences. 66(9): 1358-1359. Seo MJ, Kim SJ, Kang TH, Rim HK, Jeong HJ, Um JY, Hong SH, Kim HM. 2011. The regulatory mechanism of β-eudesmol is through the suppression of caspase-1 activation in mast cell–mediated inflammatory response. Immunopharmacology and Immunotoxicology.  33(1): 178-85. 205 Shoyama Y, Hirano H, Niskioka I. 1984. Biosynthesis of propyl cannabinoids acid and its biosynthetic relationship with pentyl and methyl cannabinoid acids. Phytochemistry. 23: 1909-1912. Small E, Beckstead HD. 1973. Cannabinoid phenotypes in Cannabis sativa. Nature. 245: 147-148. Small E, Jui PY, Lefkovitch LP. 1976. A numerical taxonomic analysis of Cannabis with special reference to species delimitation. Systematic Botany. 1(1): 67-84. Small E. 2015. Evolution and classification of Cannabis sativa (marijuana, hemp) in relation to human utilization. Botanical Review. 81: 189-294. Small E, Naraine SGC. 2016. Size matters: Evolution of large drug-secreting resin glands in elite pharmaceutical strains of Cannabis sativa (marijuana). Genetic Resources and Crop Evolution. 63(2): 349-359. Smith RN. 1975. High-pressure liquid chromatography of Cannabis identification of separated constituents. Journal of Chromatography. 115: 101-106. Smith RN, Vaughan CG. 1976. High-pressure liquid chromatography of Cannabis quantitation analysis of acidic and neutral cannabinoids. Journal of Chromatography. 129: 347-354. Smith RM. 1997. Identification of butyl cannabinoids in marijuana. Journal of Forensic Sciences. 42: 610-618. Soler S, Gramazio P, Figas MR, Vilanova S, Rosa E, Llosa ER, Borras D, Plazas M, Prohens J. 2017. Genetic structure of Cannabis sativa var. indica cultivars based on genomic SSR (gSSR) marker: Implications for breeding and germplasm management. Industrial Crops & Products. 104: 171-178. Souza RH, Cardoso MS, Menezes CT, Silva JP, De Sousa DP, Batista JS. 2011. Gastroprotective activity of α-terpineol in two experimental models of gastric ulcer in rats. Daru: Journal of Faculty of Pharmacy, Tehran University of Medical Sciences. 19(4): 277-281. 206 Stolker AAM, van Schoonhven J, de Vries AJ, Bobeldijk-Pastorova I, Vaes WHJ, van den Berg R. Determination of cannabinoids in Cannabis products using liquid chromatography–ion trap mass spectrometry. Journal of Chromatography A. 1058: 143-151. Sumner LW, Mendes P, Dixon RA. 2003. Plant metabolomics: large-scale phytochemistry in the functional genomics era. Phytochemistry. 62(6): 817-836. Swift W, Wong A, Li KM, Arnold JC, McGregor IS. 2013. Analysis of Cannabis seizures in NSW, Australia: Cannabis potency and cannabinoid profile. PLoS One. 8(7): e70052. Takemoto H, Yagura T, Ito M. 2009. Evaluation of volatile components from spikenard: valerena-4,7(11)-diene is a highly active sedative compound. Journal of Natural Medicines. 63(4): 380-385. Taura F, Morimoto S, Shoyama T, Mechoulam R. 1995. First direct evidence for the mechanism of Δ9-tetrahydrocannaninolid acid biosynthesis. Journal of the American Chemical Society. 117: 9766-9767. Taura F, Morimoto S, Shoyama T. 1996. Purification and characterization of cannabidiolic acid synthase from Cannabis sativa L. The Journal of Biological Chemistry. 271: 17411-17416. Tholl D. 2006. Terpene synthases and the regulation, diversity and biological roles of terpene metabolism. Current Opinion in Plant Biology. 9(3): 297-304. Tholl D. 2015. Biosynthesis and biological functions of terpenoids in plants. In: Schrader J., Bohlmann J. (eds) Biotechnology of Isoprenoids. Advances in Biochemical Engineering/Biotechnology. 148: 63-106. Toyota M, Shimamura T, Ishii H, Renner M, Braggins J, Asakawa Y. 2002. New bibenzyl cannabinoid from the New Zealand liverwort Radula marginata. Chemical and Pharmaceutical Bulletin. 50: 1390–1392. Tramèr MR, Carroll D, Campbell FA, Reynolds JM.  Moore RA, McQuay HJ. 2001. Cannabinoids for control of chemotherapy induced nausea and vomiting: quantitative systematic review. British Medical Journal. 232: 1-8. 207 Tsoyi K, Jang HJ, Lee YS, Kim YM, Kim HJ, Seo HG, Lee JH, Kwak JH, Lee DU, Chang KC. 2011. (+)-Nootkatone and (+)-valencene from rhizomes of Cyperus rotundus increase survival rates in septic mice due to heme oxygenase-1 induction. Journal of Ethnopharmacology. 137(3): 1311-1317. Tsuneki H, Ma EL, Kobayashi S, Sekizaki N, Maekawa K, Sasaoka T, Wang MW, Kimura I. 2005. Antiangiogenic activity of β-eudesmol in vitro and in vivo. European Journal of Pharmacology. 512: 105-115. Tugizimana F, PiaterL, Dubery, I. 2013. Plant metabolomics: A new frontier in phytochemical analysis. South African Journal of Science. 109: 1-11. Turi CE, Finley J, Shipley PR, Murch SJ, Brown PN. 2015. Metabolomics for phytochemical discovery: Development of statistical approaches using a cranberry model system. Journal of Natural Products. 78: 953-966. Turner JC, Hemphill JK, Mahlberg PG. 1978. Quantitative determination of cannabinoids in individual glandular trichomes of Cannabis sativa L (Cannabaceae). American Journal of Botany. 65(10): 1103-1106. Turner CA, ElSohly MA, Boeren EG. 1980. Constituents of Cannabis sativa L. XVII. A review of the natural constituents. Journal of Natural Products. 43(2): 169-234. Turner JC, Hemphill JK, Mahlberg PG. 1981a. Interrelationships of glandular trichomes and cannabinoid content. I: Developing pistillate bracts of Cannabis sativa L (Cannabaceae). UNODC – Bulletin on Narcotics. 2: 59-69. Turner JC, Hemphill JK, Mahlberg PG. 1981b. Interrelationships of glandular trichomes and cannabinoid content II. Developing vegetative leaves of Cannabis sativa L. (Cannabaceae). UNODC – Bulletin on Narcotics. 3: 63-71. Turner JC, Mahlberg PG. 1982. Simple high-performance liquid chromatographic method for separating acidic and neutral cannabinoids in Cannabis sativa L. Journal of Chromatography. 253: 295-303. 208 Tyagi AK, Prasad S, Yuan W, Li S, Aggarwal BB. 2015. Identification of a novel compound (β-sesquiphellandrene) from turmeric (Curcuma longa) with anticancer potential: comparison with curcumin. Investigational New Drugs. 33(6): 1175-1186. United Nations Office on Drugs and Crime. 2009. Recommended methods for the identification and analysis of Cannabis and Cannabis products. United Nations, New York. Accessed October 3, 2018 at https://www.unodc.org/documents/scientific/ST-NAR-40-Ebook_1.pdf Upton R, Craker L, ElSohly M, Romm A, Russo E, Sexton M. 2014. Cannabis Inflorescence: Cannabis spp. Standards of identity, analysis and quality control. Revision 2014. American Herbal Pharmacopeia. Scotts Valley, CA. van Bakel H, Stout JM, Cote AG, Tallon CM, Sharpe AG, Hughes TR, Page JE. 2011. The draft genome and transcriptome of Cannabis sativa. Genome Biology. 12: R102. Vanhoenacker Gmm Rompaey P, De Keukeleire D, Sandra P. 2002. Chemotaxonomic features associated with flavonoids of cannabinoid free Cannabis (Cannabis sativa subsp sativa L.) in relation to hops (Humulus lupulus L.). Natural Product Letters. 16(1): 57-63. Veress T, Szanto JI, Leisztner L. 1990. Determination of cannabinoid acids by high-performance liquid chromatography of their neutral derivatives formed by thermal decarboxylation. I. Study of the decarboxylation process in open reactors. Journal of Chromatography. 520: 339-347. Walsh Z, Callaway R, Belle-Isle L, Capler R, Kay R, Lucas P, Holtzman S. 2013. Cannabis for therapeutic purposes: Patient characteristics, access and reasons for use. International Journal of Drug Policy. 24: 511-516. Wang R, He LS, Xia B, Tong JF, Li N, Peng F. 2009. A micropropagation system for cloning of hemp (Cannabis sativa L.) by shoot tip culture. Pakistan Journal of Botany. 41(2): 603-608. Wang M, Wang YH, Avula B, Radwan MM, Wana AS, Medmedic Z, van Antwerp J, ElSohly MA, Khan IA. 2017. Quantitative determination of cannabinoids in Cannabis and Cannabis products using ultra-high performance supercritical fluid chromatography and diode array/mass spectrometric detection. Journal of Forensic Sciences. 62(3): 602-611. 209 Wang YH, Avula B, ElSohly MA, Radwan MM, Wang M, Wanas AS, Mehmedic Z, Khan IA. 2018. Quantitative determination of Δ9-THC, CBG, CBD, their acid precursors and five neutral cannabinoids by UHPLC-UV-MS. Planta Medica. 84: 260-266. Wang X, Harrington PB, Baugh SF. 2017. Comparative study of NMR spectral profiling for the characterization and authentication of Cannabis. Journal of AOAC International. 100(5): 1356-1364. Wang X, Harrington PB, Baugh SF. 2018. Effect of preprocessing high-resolution mass spectra on the pattern recognition of Cannabis, hemp and liquor. Talanta. 180(1): 229-238. Washington State Liquor and Cannabis Board. 2014. Initiative Measure No. 502.   Watts P, Long G, Meek ME. 2004. Concise International Chemical Assessment Document 58: Chloroform. In: World Health Organization. Accessed October 3, 2018 at http://www.who.int/ipcs/publications/cicad/en/cicad58.pdf Wheals 1975, Smith RN. 1975. A comparison of high-pressure liquid chromatography with other chromatographic techniques. Journal of Chromatography. 105: 396-400. Worley B, Powers R. 2013. Multivariate analysis in metabolomics. Current Metabolomics. 1(1): 92-107. Wu Y, Duan S, Zhao L, Gao Z, Luo M, Song S, Xu W, Zhang C, Ma C, Wang S. 2016. Aroma characterization based on aromatic series analysis in table grapes. Scientific Reports. 3: 31116. Yu CW, Li WH, Hsu FL, Yen PL, Chang ST, Liao VH. 2014. Essential oil alloaromadendrene from mixed-type Cinnamomum osmophloeum leaves prolongs the lifespan in Caenorhabditis elegans. Journal of Agricultural and Food Chemistry. 62(26): 6159-6165. Yun J. 2014. Limonene inhibits methamphetamine-induced locomotor activity via regulation of 5-HT neuronal function and dopamine release. Phytomedicine. 21: 883-887.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0376606/manifest

Comment

Related Items